The structural basis of Rubisco phase separation in the pyrenoid

Abstract

Approximately one-third of global CO₂ fixation occurs in a phase-separated algal organelle called the pyrenoid. The existing data suggest that the pyrenoid forms by the phase separation of the CO₂-fixing enzyme Rubisco with a linker protein; however, the molecular interactions underlying this phase separation remain unknown. Here we present the structural basis of the interactions between Rubisco and its intrinsically disordered linker protein Essential Pyrenoid Component 1 (EPYC1) in the model alga Chlamydomonas reinhardtii. We find that EPYC1 consists of five evenly spaced Rubisco-binding regions that share sequence similarity. Single-particle cryo-electron microscopy of these regions in complex with Rubisco indicates that each Rubisco holoenzyme has eight binding sites for EPYC1, one on each Rubisco small subunit. Interface mutations disrupt binding, phase separation and pyrenoid formation. Cryo-electron tomography supports a model in which EPYC1 and Rubisco form a codependent multivalent network of specific low-affinity bonds, giving the matrix liquid-like properties. Our results advance the structural and functional understanding of the phase separation underlying the pyrenoid, an organelle that plays a fundamental role in the global carbon cycle.

Extended Data Fig. 1. The EPYC1 peptides with the highest binding affinities to Rubisco were chosen for structural studies.

a, Diagram indicating the differences between the previously defined sequence repeats and the newly defined sequence repeats on full-length EPYC1. b, To verify the Rubisco-binding regions on EPYC1, surface plasmon resonance (SPR) was used to measure the binding of EPYC1 peptides to Rubisco. Purified Rubisco was immobilized on a sensor surface, and the EPYC1 peptides in solution were injected over the surface. The binding activity was recorded in real time in a sensorgram. c, The peptides used in SPR experiments are shown aligned to the sequence as shown in Fig. 1. The Rubisco-binding signal from the SPR experiment of each peptide is shown after normalization to the peptide’s molecular weight. EPYC1_49-72 (boxed in red) and EPYC1_106-135 (boxed in pink) were chosen for structural studies based on their reproducible high Rubisco binding signal. d, The Rubisco-binding response of the EPYC1_49-72 peptide at different concentrations was measured by SPR. e, The binding responses shown in (d) were fitted to estimate the K_D of EPYC1_49-72 peptide binding to Rubisco.

Extended Data Fig. 2. Single-particle cryo-EM data collection and image processing procedure.

a-c, Representative micrographs of the apo Rubisco sample (a), the Rubisco-EPYC1_49-72 complex (b) and the Rubisco-EPYC1_106-135 complex (c). Scale bars = 100 nm. d-f, Representative 2D class averages of the apo Rubisco sample (d), the Rubisco-EPYC1_49-72 complexes (e) and the Rubisco-EPYC1_106-135 complexes (f). g-i, Overview of the workflow for single-particle data processing for the apo Rubisco sample (g), the Rubisco-EPYC1_49-72 sample (h) and the Rubisco-EPYC1_106-135 sample (i). j-l, Local resolution estimation of the final refined apo Rubisco map (j), the final refined Rubisco-EPYC1_49-72 complex map (k) and the final refined Rubisco-EPYC1_106-135 complex map (l).

Extended Data Fig. 3. Cryo-EM analysis and resolution of apo Rubisco and Rubisco-EPYC1 peptide complexes in this study.

a-b, Representative cryo-EM density quality showing an α-helix of residues 214-232 in chain A (one of the Rubisco large subunits) (a) and a β-sheet of residues 36-43 in chain A (b) of the Rubisco-EPYC1_49-72 density map and structural model. The densities are shown as meshwork in gray. The backbones of the structural model are in ribbon representation, and side chains are shown in stick representation. c-d, Representative cryo-EM density quality showing water molecules as orange spheres. One water molecule between R312 and E136 on chain A is shown in panel c, and another water molecule between D137 and K316 on chain A is shown in panel d. e, Fourier shell correlation (FSC) curves of the final density maps of apo Rubisco and the Rubisco-EPYC1 peptide complexes.

Extended Data Fig. 4. Comparison of our EM structure of apo Rubisco and the published X-ray crystallography structure (1gk8) of Rubisco purified from Chlamydomonas reinhardtii, and comparison of our EM structure of apo Rubisco and Rubisco bound with EPYC1_49-72 peptide.

a, Comparison of the structure of the small subunit of apo Rubisco obtained here by EM with 1gk8. The EM structure has additional C-terminus density past residue 126, circled by a red dashed line. b, Comparison of our two EM structures of the small subunit: from apo Rubisco and from EPYC1_49-72 peptide-bound Rubisco. c, Comparison of the structure of the large subunit of apo Rubisco obtained here by EM with 1gk8. The three major differences found between the X-ray structure and the EM structure of the large subunit are circled with red dashed lines. d, Comparison of our two EM structures of the large subunit: from apo Rubisco and from EPYC1_49-72 peptide-bound Rubisco. The major difference found between the EPYC1_49-72 peptide-bound structure and the apo EM structure was the loop between K175 and L180 of the large subunit, which is shown circled by a red dashed line.

Extended Data Fig. 5. Additional residues may contribute to the interaction between EPYC1 and Rubisco.

Our Rubisco-EPYC1_49-72 peptide structure suggests that R56 of the EPYC1_49-72 peptide may interact with D31 of the Rubisco small subunit and E433 of the Rubisco large subunit (the atoms of the backbone of E433 are also shown to display the possible interaction). R51 of the EPYC1_49-72 peptide may form a salt bridge with Y32 of the Rubisco small subunit. Residues S57 and V58 of the EPYC1_49-72 peptide are close to D31 in the structure, which may explain why replacing either of these residues with a negatively charged residue disrupts binding (Fig. 4a).

Extended Data Fig. 6. The EPYC1_106-135 peptide binds to Rubisco small subunit α-helices via salt bridges and a hydrophobic pocket in a similar manner to the EPYC1_49-72 peptide.

a, The EPYC1_106-135 peptide represents the second, third and fourth Rubisco-binding regions of EPYC1 indicated by pink lines and dash line (the peptide is a perfect match to the second and fourth Rubisco-binding regions, and there is a one-amino acid difference between the peptide and the third repeat). b-c, Side view (b) and top view (c) of the density map of the EPYC1_106-135 peptide-Rubisco complex. Dashes in panel b indicate regions shown in panels d-i. d-e, Front (d) and side (e) views of the EPYC1_106-135 peptide (red) bound to the two α-helices of the Rubisco small subunit (blue). f-g, Three pairs of residues form salt bridges between the helix of the EPYC1_106-135 peptide and the helices on the Rubisco small subunit. Shown are front (f) and side (g) views as in panel d and panel e. The distances from EPYC1 K127, R134 and E129 to Rubisco small subunit E24, D23 and R91 are 2.96 Å, 3.17 Å, and 2.68 Å, respectively. h-i, A hydrophobic pocket is formed by three residues of the EPYC1_106-135 peptide and three residues of helix B of the Rubisco small subunit. Shown are front (h) and side (i) views as in panel d and panel e. j, Summary of the interactions observed between the EPYC1_106-135 peptide and the two α-helices of the Rubisco small subunit. Helices are highlighted; the residues mediating interactions are bold; salt bridges are shown as dotted lines; residues contributing to the hydrophobic pocket are shown in black. k, Color keys used in this figure.

Extended Data Fig. 7. Surface plasmon resonance analysis of binding of point mutants of EPYC1_55-72 to Rubisco.

The wild-type (WT) peptide or peptides with the indicated mutations were synthesized, and their Rubisco-binding signal was measured by surface plasmon resonance.

Extended Data Fig. 8. Interface residues on EPYC1 identified by cryo-EM are important for binding and phase separation of EPYC1 and Rubisco.

a, SDS-PAGE analysis of purified proteins used for in vitro phase separation experiments. WT = wild-type EPYC1; R/K = EPYC1^{R64A/K127A/K187A/K248A/R314}. b-c, A droplet sedimentation assay was used as a readout of phase separation complementary to the microscopy analyses shown in Fig. 4b. Proteins at indicated concentrations were mixed and incubated for 10 minutes, then condensates were pelleted by centrifugation. Supernatant (S) and pellet (P) fractions were run on a denaturing gel. The negative controls with no Rubisco or with no EPYC1 are shown in (b), and the wild-type Rubisco with wild-type EPYC1 or mutant EPYC1 are shown in (c). Data shown here are representative of two independent replicates.

Extended Data Fig. 9. Yeast two-hybrid assays of interactions between EPYC1 and wild-type or mutated Rubisco small subunit.

Colonies are shown after 3 days growth on plates. A subset of the data shown in this figure is shown in Fig. 5a.

Extended Data Fig. 10. Selection of the Rubisco small subunit mutant strains for phenotype analysis.

a, The Rubisco small subunit-less mutant T60 (Δrbcs) was transformed with DNA encoding wild-type and mutant Rubisco small subunits (RBCS) to produce candidate transformants with the genotypes Δrbcs;RBCS^WT, Δrbcs;RBCS^D23A/E24A, and Δrbcs;RBCS^M87D/V94D. Total protein extracts for three strains from each transformation were separated on a polyacrylamide gel. b, The gel shown in panel a was probed by Western blot using a polyclonal antibody mixture that detects both large and small Rubisco subunits. The experiments shown in panel a and b were performed once for selecting the candidate transformants with the highest RBCS expression level from each genotype, in case any phenotype may be caused by low expression level of Rubisco. Selected strains are indicated by an arrow below the lanes and were used for the subsequent phenotypic analyses shown in Fig. 5 and panel c. c, Additional representative TEM images of whole cells of the strains expressing wild-type, D23A/E24A, and M87D/V94D Rubisco small subunit. Scale bar = 500 nm. For each strain, at least 25 images (one image for one cell) were taken and showing similar results.

Fig. 1 |. EPYC1 consists of five tandem sequence repeats, each of which contains a Rubisco-binding region.

a, A representative (N=15) transmission electron microscopy (TEM) image of a Chlamydomonas cell. Scale bar = 1 μm. b, Cartoon depicting the chloroplast and pyrenoid in the image shown in panel a. The blue dots indicate the location of Rubisco enzymes clustered in the pyrenoid matrix. c, We hypothesized that pyrenoid matrix formation is mediated by multivalent interactions between Rubisco and the intrinsically disordered protein EPYC1. d, We designed an array of 18 amino acid peptides tiling across the full length EPYC1 sequence. e, Incubation of the array with purified Rubisco allows identification of peptides that bind to Rubisco. f, Image of the Rubisco binding signal from the peptide tiling array. g, The Rubisco binding signal was quantified and plotted for each peptide as a function of the position of the middle of the peptide along the EPYC1 sequence. The initial 26 amino acids of EPYC1 correspond to a chloroplast targeting peptide (cTP), which is not present in the mature protein. Results are representative of three independent experiments. h, The positions of EPYC1’s five sequence repeats are shown to scale with panel g. Predicted α-helical regions are shown as wavy lines. i, Primary sequence of EPYC1, with the five sequence repeats aligned. In panels h and i, the regions represented by peptides subsequently used for structural studies are underlined with red lines (EPYC1_49-72) and pink lines (EPYC1_106-135). EPYC1_106-135 is an exact match to the underlined sequence of Repeats 2 and 4, and has a one-amino acid difference from the corresponding region in Repeat 3 (dashed underline).

Fig. 2 |. EPYC1 binds to Rubisco small subunits.

a, Peptide EPYC1_49-72, corresponding to the first Rubisco-binding region of EPYC1, was incubated at saturating concentrations with Rubisco prior to single-particle cryo-electron microscopy. b-e, Density maps (b, d) and cartoons (c, e) illustrate the side views (b, c) and top views (d, e) of the density map of the EPYC1 peptide-Rubisco complex. Dashed boxes in panel b indicate regions shown in Fig. 3a–f.

Fig. 3 |. EPYC1 binds to Rubisco small subunit α-helices via salt bridges and a hydrophobic pocket.

a-b, Front (a) and side (b) views of the EPYC1_49-72 peptide (red) bound to the two α-helices of the Rubisco small subunit (blue). c-d, Three pairs of residues form salt bridges between the helix of the EPYC1_49-72 peptide and the helices on the Rubisco small subunit. Shown are front (c) and side (d) views as in panel a and panel b. The distances from EPYC1 R64, R71 and E66 to Rubisco small subunit E24, D23 and R91 are 3.06 Å, 2.78 Å, and 2.79 Å, respectively. e-f, A hydrophobic pocket is formed by three residues of the EPYC1_49–72 peptide and three residues of helix B of the Rubisco small subunit. Shown are front (e) and side (f) views as in panel a and panel b. g, Summary of the interactions observed between the EPYC1_49-72 peptide and the two α-helices of the Rubisco small subunit. Helices are highlighted; the residues mediating interactions are bold; salt bridges are shown as dotted lines; residues contributing to the hydrophobic pocket are shown in black.

Fig. 4 |. Interface residues on EPYC1 are required for binding and phase separation of EPYC1 and Rubisco in vitro.

a, Rubisco binding to a peptide array representing every possible single amino acid substitution for amino acids 56-71 of EPYC1. The binding signal was normalized by the binding signal of the original sequence. b, The effect of mutating the central R or K in each of EPYC1’s Rubisco-binding regions on in vitro phase separation of EPYC1 with Rubisco. Scale bar = 10 μm. For each condition, the experiment was performed twice independently with similar results.

Fig. 5 |. Interface residues on Rubisco are required for yeast two-hybrid interactions between EPYC1 and Rubisco, and for pyrenoid matrix formation in vivo.

a, The importance of Rubisco small subunit residues for interaction with EPYC1 was tested by mutagenesis in a yeast two-hybrid experiment. b, The Rubisco small subunit-less mutant T60 (Δrbcs) was transformed with wild-type, D23A/E24A or M87D/V94D Rubisco small subunits. Serial 1:10 dilutions of cell cultures were spotted on TP minimal medium and grown in air or 3% CO₂. c-d, Representative electron micrographs of whole cells (c) and corresponding pyrenoids (d) of the strains expressing wild-type, D23A/E24A, and M87D/V94D Rubisco small subunit. Dashes in panel c indicate regions shown in panel d. Scale bars = 500 nm. At least 25 cells were imaged for each strain; additional representative images are shown in Extended Data Fig. 10c.

Fig. 6 |. A model for matrix structure consistent with in situ Rubisco positions and orientations.

a, The pyrenoid matrix was imaged by cryo-electron tomography. An individual slice through the three-dimensional volume is shown. Scale bar = 200 nm. b, The positions and orientations of individual Rubisco holoenzymes (blue) were determined with high sensitivity and specificity (97.5% positive identification) by template matching, subtomogram averaging, and classification and then mapped back into the tomogram volume shown in panel a. The yellow box indicates the region shown in panel c. Scale bar = 200 nm. c, The distances (yellow) between the nearest EPYC1-binding sites (red) on neighboring Rubisco holoenzymes (blue) were measured. The view is from inside the matrix; in some cases the nearest EPYC1 binding site is on a Rubisco that is out of the field of view, causing some yellow lines to appear unconnected in this image. Scale bar = 20 nm. The data shown in panels a-c are representative of the five independent tomograms used for this study. d, Histogram showing the distances between the nearest EPYC1 binding sites on neighboring Rubisco holoenzymes. The black line indicates the median, and the yellow shading indicates 95% confidence interval based on data from five independent tomograms. The estimated energy required for stretching a chain of 40 amino acids a given distance is shown in blue. e, A 3D model illustrates how EPYC1 (red) could crosslink multiple Rubisco holoenzymes (blue) to form the pyrenoid matrix. The conformations of the intrinsically disordered linkers between EPYC1 binding sites were modeled hypothetically.