The Eukaryotic CO2-Concentrating Organelle Is Liquid-like and Exhibits Dynamic Reorganization

Abstract

Approximately 30%-40% of global CO₂ fixation occurs inside a non-membrane-bound organelle called the pyrenoid, which is found within the chloroplasts of most eukaryotic algae. The pyrenoid matrix is densely packed with the CO₂-fixing enzyme Rubisco and is thought to be a crystalline or amorphous solid. Here, we show that the pyrenoid matrix of the unicellular alga Chlamydomonas reinhardtii is not crystalline but behaves as a liquid that dissolves and condenses during cell division. Furthermore, we show that new pyrenoids are formed both by fission and de novo assembly. Our modeling predicts the existence of a "magic number" effect associated with special, highly stable heterocomplexes that influences phase separation in liquid-like organelles. This view of the pyrenoid matrix as a phase-separated compartment provides a paradigm for understanding its structure, biogenesis, and regulation. More broadly, our findings expand our understanding of the principles that govern the architecture and inheritance of liquid-like organelles.

Keywords: CO(2) concentrating mechanism; Chlamydomonas reinhardtii; Rubisco; biological phase transitions; carbon fixation; cryo-electron tomography; liquid-like organelles; magic numbers; organelle inheritance; pyrenoid.

Figure 1. The Pyrenoid is Inherited Primarily by Fission

(A and B) Confocal Z-sum images of pyrenoid divisions by fission, with chlorophyll autofluorescence shown in magenta, and RbcS1-Venus in green; t=0 is the first observation of a gap in chlorophyll between the daughter pyrenoids in the first division shown. Dashed curves represent approximate chloroplast outlines in the mother (white) and daughter (pink) cells. (C) A cartoon of the approximate locations of the pyrenoid (green), chloroplast (magenta), and cell membrane (black outline). (D) Example of pyrenoid fission in EPYC1-Venus, annotated as in (B). (E) Example of the progressing chloroplast cleavage furrow (arrows) appearing to separate daughter pyrenoids. Images are 2D snapshots of 3D Z-stack reconstructions. (F) Average and standard deviation of the durations of chloroplast (magenta) and pyrenoid (green) fissions in RbcS1-Venus (left; n = 28 1^st and 2^nd divisions) and EPYC1-Venus (right; n = 22 1^st and 25 2^nd divisions). (G and H) Duration and relative timing of chloroplast (magenta) and pyrenoid (green) division for the pyrenoid fissions plotted in (F). Each bar represents a different division. See also Figures S1–2, and Movie S1.

Figure 2. Pyrenoids Can Also be Inherited by Other Means

(A – D) Examples of other types of pyrenoid inheritance patterns observed in RbcS1-Venus (A, B) and EPYC1-Venus (C, D) cells. (A, C) One daughter (blue) inherits an entire pyrenoid from the mother cell (white) and another daughter (orange) inherits neither a pyrenoid nor puncta. (B, D) One daughter (blue) inherits the entire pyrenoid, and puncta appear in the other daughter (yellow) and coalesce into a new pyrenoid. (E – F) Proportion of observed RbcS1-Venus (E) and EPYC1-Venus (F) daughter cells that exhibited each observed inheritance pattern; the distribution of inheritance patterns in EPYC1-Venus cells was not significantly different from that of RbcS1-Venus cells (Chi-square test, p = 0.8). (G – I) Stills from time course image captures in which pyrenoids were observed to grow or coalesce from puncta that appeared in the chloroplast stroma during division. (G – H) RbcS1-Venus; (I) EPYC1-Venus. Images are 2D projections of the sum of pixel values in each channel in a Z-stack through the whole cell at each time point. The chloroplast of the dividing cell of interest in each series is outlined in white. Arrows point to growing or coalescing pyrenoids. t=0 is defined as the first minute at which the daughter chloroplasts are observed to be distinct in 3D. See also Figures S1–2, and Movies S2–3.

Figure 3. A “Bridge” of Matrix Material Connects Nascent Daughter Pyrenoids During Fission

(A – E) Examples of pyrenoid fissions in five RbcS1-Venus cells. Magenta is chlorophyll autofluorescence; green is RbcS1-Venus. (F) Example of pyrenoid fission in an EPYC1-Venus cell. Magenta is chlorophyll autofluorescence; green is EPYC1-Venus. Images are 2D projections of the sum of pixel values in each channel from a Z-stack through the whole cell. See also Movie S1.

Figure 4. The Pyrenoid Matrix is not Crystalline but Exhibits Short-Range Liquid-Like Order

(A) Slice through a tomographic volume of the native Chlamydomonas pyrenoid. (B) Segmentation of the tomogram shown in (A) with localized positions of 46,567 Rubisco holoenzymes (magenta) mapped into the volume. Green and yellow: pyrenoid tubule membranes. (C) In situ subtomogram average of Rubisco (16.5 Å resolution; Figure S4A) generated from 30,000 particles extracted from the tomogram shown in (A). (D) The local density of neighbor Rubisco particles as a function of the distance from the reference particle. Each line represents a separate tomogram, showing the sum of the local densities around every Rubisco. The distances to peaks of high local Rubisco concentration are indicated. (E) Histogram of distances from reference particles to their nearest neighbors (NN), summed from all five tomograms. Red dashed line: Gaussian distribution fit to the 13.9 nm NN peak. Light blue bars: distance to the 12 NN within 1 standard deviation (<1 SD) of the 13.9 nm peak, dark blue bars: distance to the 12 NN beyond 1 standard deviation (>1 SD) from the 13.9 nm peak, grey bars: distance to further (13+ NN) neighbors. Mean distance to the 12 NN = 15.9 nm. Inset: distribution of the number of neighbors per reference particle (mean = 4.4 neighbors) that are <1 SD from the 13.9 nm peak. (F) The normalized local density of neighbor particles (local density divided by the global density), showing the mean value ± 99% CI of the experimental data (black) compared to crystalline simulated data generated within the same tomogram volumes (Figure S4E,F): crystal structure packing (Taylor et al., 2001) (red), 13.9 nm-spaced HCP (blue). (G) The mean value ± 99% CI of the experimental data’s normalized local density (black) fit with the radial distribution function of a Lennard-Jones fluid generated by an analytical model (Morsali et al., 2005) (red) and by molecular dynamics simulations (Plimpton, 1995) (blue). (H) The normalized local density of neighbor particles, showing the mean value ± 99% CI of the experimental data (black) compared to random simulated data generated within the same tomogram volumes (Figure S4G–J): single particles (red), pairs linked by 13.9 ± 1.5 nm (yellow), and linked networks (blue). See also Figures S3–S4, and Movie S4.

Figure 5. Pyrenoid Matrix Components Mix Internally

(A and B) FRAP in live (A) and fixed (B) RbcS1-Venus pyrenoids. Cartoons depict the approximate bleached region (dark gray). Different intensity display scales are used in the pre- and post-bleach image sets. (C – D) Kymographs of the pyrenoids shown in parts (A-B), respectively. From left to right: pyrenoid cartoon showing the region used to create the kymographs (dashed rectangle); the pre- bleach section of the kymographs; and the post-bleach kymographs. (E – F) Fluorescence recovery occurs from within the pyrenoid in live pyrenoids (E), but does not occur in fixed pyrenoids (F). The x-axis is μm along the dashed regions in (C-D). (G) Average fluorescence recovery profiles ± SEM for pyrenoids in live RbcS1-Venus (blue), RCA1-Venus (red), or EPYC1-Venus (yellow) cells, and in fixed RbcS1-Venus cells (gray). (H) Average recovery rates ± SEM over the first 12 seconds in (G). ** p < 0.005; * p < 0.05 (one-way ANOVA & post-hoc Bonferroni means comparison;). (I– L) Examples of half-pyrenoid FRAP in live RCA-Venus (I, K) and EPYC1-Venus (J, L) cells, with images from the recovery time-courses (I, J) and corresponding kymographs (K, L) as shown in (A – D). See also Figure S5 and Movie S5.

Figure 6. The Pyrenoid Matrix Disperses and Re-Aggregates During Division

(A) Heat maps of the RbcS1-Venus signal during the divisions in Figure 1A-C. Times are defined as in Figure 1. (B) Raw signal from (A) plotted over time by regions of interest, representing the sum through the whole Z-stack in each masked region over time. Times of pyrenoid divisions are highlighted in gray. (C) The average signal in the pyrenoid during division is significantly lower than that 15 minutes later (Wilcoxon Matched-Pairs Signed-Ranks Test; * p ≤ 1.50⁻⁶; n = 31), shown ± SEM. (D) Timeline of an average cell division with pyrenoid fission. Chloroplast division (magenta), pyrenoid dissolution (cyan), and pyrenoid fission (green) are displayed relative to the moment the chloroplast division furrow passes between the daughter pyrenoids (t=0). Cartoons depicting the aggregation state of the pyrenoid matrix are shown above each stage, with the chloroplast outlined in black, aggregated matrix components shown as filled black circles, and partially dispersed matrix components as speckles. See also Figure S6, and Movies S1–S3.

Figure 7. Simulations of an EPYC1-Rubisco System Reveal an Effect of Binding Site Stoichiometry on the Aggregation State

(A – C), Snapshots of simulations with 3 (A), 4 (B), and 5 (C) Rubisco binding sites on EPYC1. “Rubiscos” (blue rectangles) and “EPYC1s” (red polymers) bind when they occupy the same sites in a 2D grid. Snapshots are from simulations with 10 k_BT specific bonds and 0.1 k_BT nearest-neighbor non-specific bonds. (D – F) Heat maps of the fraction of Rubiscos that are in clusters of >10 Rubiscos connected by EPYC1s with 3 (D), 4 (E), or 5 (F) Rubisco binding sites. The fraction of grid sites occupied by Rubiscos (x-axis) is varied, with an equal fraction of grid sites occupied by EPYC1s. The specific bond energy (y-axis) is varied, while the nonspecific bond energy is fixed at 0.1 k_BT. Red dots indicate the parameters used for the snapshots in (A – C). (G – I), Snapshots of off-lattice 3D simulations with 3 (G), 4 (H), and 5 (I) Rubisco binding siteson EPYC1 for Rubisco. “Rubiscos” (blue spheres with 4 binding sites on each end) and “EPYC1s” (red polymers) bind when their binding sites overlap. Inset in (I): zoom-in with one Rubisco and one EPYC1 with 5 binding sites; 4 of the 5 binding sites of the EPYC1 are in specific bonds with the Rubisco. Snapshots are from simulations with 10 k_BT specific bonds and a Lennard-Jones nonspecific interaction with ε = 0.1 k_BT. The molecules occupy ~2% of the total space in these simulations, with equal total numbers of EPYC1 and Rubisco binding sites. (J) Fraction of Rubiscos in clusters of >10 Rubiscos for EPYC1s with different numbers of binding sites, in the 2D model. The specific bond energy is 10 k_BT and the nonspecific bond energy is 0.1 k_BT. (K) The concentration of Rubisco at which clustering begins for systems with different numbers of EPYC1 binding sites in the 2D model. The onset is determined from the curves in (J) (see Figure S7). (L) Heat map of the distribution of cluster sizes for different numbers of EPYC1 binding sites in the 2D model. Each column depicts the normalized cluster-size distribution at [Rubisco] = 0.15, with 10 k_BT specific bond energy and 0.1 k_BT nonspecific bond energy. (M) Schematic of a possible mechanism by which magic numbers could regulate the formation and dissolution of pyrenoids: EPYC1s with <4 binding sites favor Rubisco clustering in the pyrenoid (left), while EPYC1s with <4 binding sites form stable 2:1 complexes of EPYC1:Rubisco that dissolve into the chloroplast stroma (right). See also Figure S7, Table S1, and Movie S6.