A systematic exploration of bacterial form I rubisco maximal carboxylation rates

Abstract

Autotrophy is the basis for complex life on Earth. Central to this process is rubisco-the enzyme that catalyzes almost all carbon fixation on the planet. Yet, with only a small fraction of rubisco diversity kinetically characterized so far, the underlying biological factors driving the evolution of fast rubiscos in nature remain unclear. We conducted a high-throughput kinetic characterization of over 100 bacterial form I rubiscos, the most ubiquitous group of rubisco sequences in nature, to uncover the determinants of rubisco's carboxylation velocity. We show that the presence of a carboxysome CO₂ concentrating mechanism correlates with faster rubiscos with a median fivefold higher rate. In contrast to prior studies, we find that rubiscos originating from α-cyanobacteria exhibit the highest carboxylation rates among form I enzymes (≈10 s^-1 median versus <7 s^-1 in other groups). Our study systematically reveals biological and environmental properties associated with kinetic variation across rubiscos from nature.

Keywords: Carbon Fixation; Carboxysome; Cyanobacteria; Photosynthesis; Rubisco.

Figure 1. Systematic exploration of the diversity of bacterial form I rubisco.

(A) Number of bacterial form I rubisco variants with a carboxylation rate reported across the literature and in this study. (B) Histogram of the carboxylation rates measured in this study and across the literature. Source data are available online for this figure.

Figure 2. Large-scale analysis of the biological parameters associated with fast carboxylating rubiscos.

(A–C) Box and cumulative distribution plots of rubisco carboxylation rates from different clusters: (A) chemo- and phototrophic bacterial rubiscos, (B) carboxysome-associated rubiscos and their counterparts, (C) α- and β-cyanobacterial, and carboxysome-associated proteobacterial rubiscos. To ensure unbiased study of every group, we selected n = 19 and 9 class-representative chemo- and phototroph-associated rubiscos, n = 9 and 9 class-representative carboxysome-associated and non-associated rubiscos, and n = 15, 19, and 20 class-representative α- and β- cyanobacterial, and carboxysome-associated proteobacterial rubiscos respectively (see “Methods”). Boxes represent the first and third quartiles, with the medians indicated by the central lines. Whiskers extend to the lowest and highest values within 1.5 times the interquartile range. Mann–Whitney U test (A, B) or Kruskal–Wallis followed by Dunn multiple comparison tests (C) were applied. **P < 0.01, ***P < 0.001. Legend abbreviations are as follows: α-cyano, α-cyanobacterial rubisco; β-cyano, β-cyanobacterial rubisco; CCM-proteo, carboxysome-associated proteobacterial rubisco. Source data are available online for this figure.

Figure 3. The presence of a carboxysome is the primary factor influencing form I rubisco carboxylation rate.

Feature importance was determined using absolute SHAP (Shapley additive explanations) values from a random forest regressor model. The model assessed the rubisco carboxylation rate based on bacterial trophic mode, carboxysome-association, and belonging to specific carboxysome-expressing bacterial group: alpha-cyanobacteria, beta-cyanobacteria, or carboxysome-associated proteobacteria (CCM-proteobacteria). Error bars are the standard deviations across 100 different train-test splits.

Figure EV1. Sequential screening strategy used for the selection of rubisco variants for characterization.

An iterative clustering approach was employed to screen the totality and specific subgroups of the form I rubisco family. The number of active rubiscos and measured carboxylation rates are indicated at each step.

Figure EV2. Carboxylation rates measured in this study have similar ranking as rates published in the literature in spite of different techniques and conditions.

Values in both axes are not supposed to be equal as those measured in this study result from coupled assays, which tend to underestimate the rates compared to direct assays used in the literature, and on the other side they were performed at 30 °C, resulting in faster rates compared to literature measurements done at 25 °C.

Figure EV3. Random forests modeling of form I rubisco carboxylation rate as a function of the main parameters from this study.

(A) Examples of decision trees out of the hundred trained in the random forests. (B) Measured against predicted carboxylation rates for the three most influential features according to the model. Each dot is color-coded based on the value of the respective feature (yellow/black: phototrophic/chemotrophic; green/black: carboxysome/non-carboxysome associated; cyan/black: α-cyanobacterial/non-α-cyanobacterial rubisco). RMSE = 2.1 s⁻¹; average explained variance score = 0.55.