Trajectories for the evolution of bacterial CO2-concentrating mechanisms

Abstract

Cyanobacteria rely on CO₂-concentrating mechanisms (CCMs) to grow in today's atmosphere (0.04% CO₂). These complex physiological adaptations require ≈15 genes to produce two types of protein complexes: inorganic carbon (Ci) transporters and 100+ nm carboxysome compartments that encapsulate rubisco with a carbonic anhydrase (CA) enzyme. Mutations disrupting any of these genes prohibit growth in ambient air. If any plausible ancestral form-i.e., lacking a single gene-cannot grow, how did the CCM evolve? Here, we test the hypothesis that evolution of the bacterial CCM was "catalyzed" by historically high CO₂ levels that decreased over geologic time. Using an E. coli reconstitution of a bacterial CCM, we constructed strains lacking one or more CCM components and evaluated their growth across CO₂ concentrations. We expected these experiments to demonstrate the importance of the carboxysome. Instead, we found that partial CCMs expressing CA or Ci uptake genes grew better than controls in intermediate CO₂ levels (≈1%) and observed similar phenotypes in two autotrophic bacteria, Halothiobacillus neapolitanus and Cupriavidus necator. To understand how CA and Ci uptake improve growth, we model autotrophy as colimited by CO₂ and HCO₃^-, as both are required to produce biomass. Our experiments and model delineated a viable trajectory for CCM evolution where decreasing atmospheric CO₂ induces an HCO₃^- deficiency that is alleviated by acquisition of CA or Ci uptake, thereby enabling the emergence of a modern CCM. This work underscores the importance of considering physiology and environmental context when studying the evolution of biological complexity.

Keywords: Earth history; carbon fixation; evolution; photosynthesis; synthetic biology.

Fig. 1.

Mechanism and potential routes for the evolution of the bacterial CO₂-concentrating mechanism. (A) Today, the bacterial CCM functions through the concerted action of three primary features - (i) an inorganic carbon (Ci) transporter at the cell membrane, and (ii) a properly-formed carboxysome structure (iii) co-encapsulating rubisco with carbonic anhydrase (CA). Ci uptake leads to a high intracellular HCO₃⁻ concentration, well above equilibrium with the external environment. Elevated HCO₃⁻ is converted to a high carboxysomal CO₂ concentration by CA activity located only there, which promotes carboxylation by rubisco. (B) Mutants lacking genes coding for essential CCM components grow in elevated CO₂ but fail to grow in ambient air, as shown here for mutations to the α-carboxysome in the proteobacterial chemoautotroph H. neapolitanus. Strains lacking the carboxysomal CA (ΔcsosCA) or an unstructured protein required for carboxysome formation (Δcsos2) failed to grow in ambient air, but grew robustly in 5% CO₂ (>10⁸ colony-forming units/ml, SI Appendix, Fig. S1). See SI Appendix, Table S4 for description of mutant strains. (C) We consider the CCM to be composed of three functionalities beyond rubisco itself: a CA enzyme (magenta), a Ci transporter (dark brown), and carboxysome encapsulation of rubisco with CA (light brown). If CO₂ levels were sufficiently high, primordial CO₂-fixing bacteria would not have needed a CCM. We sought to discriminate experimentally between the six sequential trajectories (dashed arrows) in which CCM components could have been acquired.

Fig. 2.

H. neapolitanus CCM genes contribute to growth even in super-ambient CO₂ concentrations. H. neapolitanus is a chemoautotroph that natively utilizes a CCM in low CO₂ environments. We profiled the contributions of CCM genes to autotrophic growth across a range of CO₂ levels by assaying a barcoded transposon mutant library (20) in sequencing-based batch culture competition assays (Methods). Mutational effects were estimated as the log₂ ratio of strain counts between the end point sample (e.g., 0.5% CO₂) and the 5% CO₂ preculture for each barcoded mutant (20, 28). As the library contained an average of ≈40 mutants per CCM gene, each point in (A) gives the average effect of multiple distinct mutants to a single CCM gene in a given CO₂ condition. A value of log₂(n/n₀) = −2 therefore indicates that gene disruption was, on average, associated with a fourfold decrease in mutant abundance as measured by Illumina sequencing of mutant strain barcodes (20, 28). Negative values indicate that the gene, e.g., the carboxysomal CA, contributes positively to the growth of wild-type H. neapolitanus in the given condition. Observed log₂(n/n₀) values indicated that many H. neapolitanus CCM genes contribute to growth in super-ambient CO₂ concentrations, including genes coding for Ci uptake, carboxysome shell proteins and the carboxysomal CA. Biological replicates are indicated by shading, and the gray bar gives the interquartile range of fitness effects for all ≈1700 mutants across all CO₂ levels (−0.15-0.065). Replicates were highly concordant (SI Appendix, Fig. S2). “Ci transport mutants” include 4 DAB genes in two operons (20), “carboxysome” includes 6 nonenzymatic carboxysome genes, “rubisco mutants” denote the two subunits of the carboxysomal rubisco, and “CA mutants” denote the carboxysomal CA gene csosCA. H. neapolitanus also expresses a secondary rubisco, which explains why disruption of the carboxysomal rubisco is not lethal in high CO₂ (SI Appendix, Fig. S3). Panel (B) gives the average mutational effect of CCM genes as a function of the CO₂ concentration. Negative average values in 0.5-1.5% CO₂ highlight the positive contribution of CCM genes to growth in these conditions. See SI Appendix, Figs. S2 and S3 for analysis of reproducibility and SI Appendix, Tables S1–S3 for detailed description of genes.

Fig. 3.

Expression of carboxysome genes without other CCM components does not improve the growth of an engineered rubisco-dependent E. coli in any CO₂ level tested. We recently reconstituted a functional H. neapolitanus α-carboxysome CCM in a rubisco-dependent E. coli strain, CCMB1, by expressing 20 CCM genes from 2 plasmids (2). Here, we generated plasmid variants to test whether carboxysome expression improves rubisco-dependent growth in any of the four CO₂ partial pressures during growth in a gas-controlled plate reader (Methods). Each diamond gives the end point optical density (600 nm) after 4 d of cultivation for one technical replicate of four biological replicates. The CCMB1 strain grows in elevated CO₂ (1.5 and 5%) when rubisco is expressed (“Rubisco Alone”, Left). As previously reported, expressing the full complement of CCM genes from the pCB’ and pCCM’ plasmids (“Full CCM”) enabled growth in all CO₂ levels. By replacing pCCM’ with a vector control and making an inactivating mutation to the carboxysomal CA (CsosCA C173S), we were able to express rubisco in a carboxysome without CA or Ci transport activities (“Encap. Rub.”). This strain grew similarly to the reference “Rubisco Alone” in all conditions. When the CA active site was left intact (“En. Rub.+CA”), growth improved above the “Rubisco Alone” baseline in 0.5% and 1.5% CO₂. A negative control strain carrying inactive rubisco (“En. Rub.⁻”, CbbL K194M) failed to grow in all CO₂ conditions. (B) Focusing on the growth in 1.5% CO₂ highlights the contribution of CA activity to rubisco-dependent growth. See SI Appendix, Tables S4 and S5 for description of strains and plasmids, and see SI Appendix, Figs. S4 and S5 for growth curves and analysis of statistical significance.

Fig. 4.

Expression of CA or Ci transport improves rubisco-dependent growth of CCMB1 E. coli in intermediate CO₂ levels even in the absence of other CCM components. As shown in Fig. 3, CCMB1 E. coli strains were grown for 4 d in minimal medium in a gas-controlled plate reader. Each diamond gives the optical density (600 nm) after 4 d for one technical replicate of four biological replicates (Methods). The reference strain (“Rubisco Alone”) constitutively expresses rubisco from the p1Ac plasmid and carries a second plasmid, pFA-sfGFP, as a vector control. The reference grew only in 1.5% and 5% CO₂. The remaining strains expressed both the E. coli native β CA (can) and the DAB2 Ci transporter from a pFA-family plasmid. These two activities were isolated by means of active site mutations. The negative control “+DAB^-+CA^-” expressed inactive Can (C48A, D50A) and DAB2 (DabA2 C539A) and grew less robustly than the reference in all conditions. If either active site was left intact (“+DAB+CA^-” or “+DAB^-+CA”), we observed a sizable growth improvement in both 0.5 and 1.5% CO₂. Contrary to our expectations, this growth improvement remained when both active sites were left intact (“+DAB+CA”). Panel (B) emphasizes this effect by focusing on growth in 1.5% CO₂. See SI Appendix, Table S4 for strain genotypes and SI Appendix, Figs. S6 and S7 for growth curves and analysis of statistical significance.

Fig. 5.

C. necator requires CA or Ci uptake for robust autotrophic growth in 0.5% and 1.5% CO₂. C. necator strains were grown autotrophically in minimal medium at a variety of CO₂ levels, and end point optical density was measured after 48 h (Methods). (A) Growth of the C. necator double CA knockout (ΔCA) was greatly impaired in 0.5% and 1.5% CO₂. Compared to wild-type C. necator (WT), which grew to a final OD600 of 0.73 ± 0.28 in 1.5% CO₂ (six biological replicates), growth of ΔCA was greatly impaired, reaching a final OD of 0.23 ± 0.17 (three biological replicates). Expression of either the human CA II (ΔCA+CA) or the DAB2 Ci transporter from H. neapolitanus (ΔCA+DAB) recovered robust growth which exceeded even the wild type, indicating that the wild type may not express saturating levels of CA. Panel (B) focuses on 1.5% CO₂. See SI Appendix, Fig. S9 for statistical analysis.

Fig. 6.

Coexpression of CA and Ci uptake enabled rubisco-dependent growth of CCMB1 E. coli in ambient air. Inspecting the ambient CO₂ growth data presented in Fig. 4 revealed that coexpression of CA and Ci transport (“Rub.+DAB+CA”) substantially improved rubisco-dependent growth of CCMB1 E. coli in ambient CO₂ concentrations. This effect was modest (≈0.1 OD units above the “Rubisco Alone” control) but reproducible, as indicated by end point data plotted on the inset. Curves are colored to match labels on the inset. See SI Appendix, Fig. S10 for statistical analysis.

Fig. 7.

Colimitation of autotrophic growth by CO₂− and HCO₃⁻-dependent carboxylation reactions can explain the growth improvements associated with expressing CAs and Ci transporters. (A) In autotrophs using the CBB cycle, nearly all biomass carbon derives from rubisco-catalyzed CO₂ fixation. However, autotrophs also require HCO₃⁻ for carboxylation reactions in lipid, nucleic acid, and arginine biosynthesis (–51). We expressed this diagram as a mathematical model, which we applied to understand why CA and Ci uptake improved rubisco-dependent growth. (B) The model exhibited two regimes: one wherein growth was limited by rubisco flux and another where it was limited by HCO₃⁻-dependent carboxylation (“bicarboxylation”) flux. At low rubisco levels (lighter-colored lines), growth was rubisco limited: increased rubisco activity produced faster growth, but the growth rate was insensitive to CA activity because slow spontaneous CO₂ hydration provided sufficient HCO₃⁻ to keep pace with rubisco. At higher rubisco levels (maroon lines), growth was bicarboxylation limited and increased CA activity was required for increasing rubisco activity to translate into faster growth. Increasing Ci uptake led to similar effects (SI Appendix, Fig. S12). In panel (C), color indicates the ratio of total Ci leakage (J_L,tot = J_L,C + J_L,H) to biomass production flux (J_B) at fixed rubisco activity across wide ranges of CA and Ci uptake activities. J_L,tot was calculated as the sum of CO₂ and HCO₃⁻ leakage rates (J_L,C + J_L,H) with J_L,C ≫ J_L,H in most conditions due to the much greater membrane permeability of CO₂. The so-called futile cycling, where leakage greatly exceeds biomass production (J_L,tot / J_B ≫ 1), occurs when CA and Ci uptake are coexpressed at extreme levels (redder colors). See SI Appendix for detailed description of the colimitation model.