Timing the Evolution of Cyanobacterial Antioxidants: Superoxide Dismutases


 The ancestors of cyanobacteria generated Earth’s first biogenic molecular oxygen but how they dealt with its toxicity remains unconstrained. Here we investigated when superoxide dismutase enzymes (SODs) capable of removing superoxide free radicals evolved. We found phylogenetic evidence that ancestral cyanobacteria used SODs with copper and zinc cofactors (CuZnSOD) during the Archaean. By the Paleoproterozoic, they became genetically capable of using iron, nickel, and manganese as cofactors (FeSOD, NiSOD, and MnSOD respectively). The evolution of NiSOD is particularly intriguing because it has been previously hypothesized that declining seawater Ni concentrations at the end of the Archaean caused a fundamental shift in the marine biosphere away from methanogenesis towards oxygenic photosynthesis. Our novel analyses of enzymes dealing with O2 toxicity now demonstrate that the beneficiaries of this chemical change - marine planktonic cyanobacteria - were able to utilize the remaining Ni from seawater 0.9-0.8 Ga to supplement their existing metabolic capabilities.


32
Oxygen is essential for complex life forms as it is used during aerobic respiration to create more 33 energy per mol of substrate than other available electron acceptors 1 . While today the Earth's 34 atmosphere contains ~ 21% oxygen (O2), it was at least 100,000 times lower in the Archaean (4.0 to 2.5 35 Ga) 2,3 . Just how and when O2 first appeared as a byproduct of biological evolution -oxygenic 36 photosynthesis -remains controversial, with estimates ranging from 3.8 billion years ago (Ga) 4 to 37 immediately preceding the Great Oxidation Event (GOE) 5 , which is estimated to have begun by 2.45 Ga 38 6 . Since the O2 produced was novel and highly reactive, early cyanobacteria -the first producers of O2 -39 would have found exposure to it highly toxic. Environmental pressures likely provided impetus for the 40 evolution of protective enzymes that prevented oxidative damage from reactive oxygen species (ROS)    Appendix, Fig. S1). By contrast, sodC was absent from the latter two phyla, but present in more genomes 100 (5723) of Betaproteobacteria, Firmicutes and the six previous (SI Appendix, Fig. S1). Together, sodA 101 and sodB are more widespread than sodC and sodN combined, being present in 13,748 bacterial species 102 across all ten phyla mentioned previously, as well as Fibrobacteres, Chlorobi, and Vampirovibrionia (SI

104
Cyanobacterial SOD Diversity 105 A more specific search amongst cyanobacteria revealed that most strains with sodN (51 of 55 106 strains) live in saltwater habitats (Fig. 1). They include representatives from all major clades of marine 107 taxa (Fig. 2). Ten lack a gene (named sodX) encoding NiSOD's maturation protease (SI Appendix, Table   108 S1). Of these, six contain genes encoding other SOD isoforms, while the remaining four are all 109 picocyanobacteria, i.e., all Prochlorococcus and some Synechococcus species (Fig. 2). 110 endosymbionts living in larger marine tunicates, algae, and sponges (e.g., Prochloron spp., UCYNA and 114 two strains of Synechococcus spongiarum). The 70 cyanobacteria which only use FeSOD and MnSOD 115 live in a variety of marine, terrestrial, and freshwater habitats (Fig. 2). Only three percent of cyanobacteria 116 (five of 149 strains discounting plastids) have genes encoding every SOD isoform as well as the NiSOD 117 maturation protease (Fig. 2). Paralogues of sodC were found in three genomes and paralogues of sodA 118 and/or sodB were found in at least 13 genomes (SI Appendix, Table S1).

119
The resources needed to make each SOD isoform vary. NiSOD is composed from a mode of 157 120 amino acids (range 145-166), whereas CuZnSOD is made from 177 (range 103-236) and

142
Genes encoding CuZnSOD may have been present in the shared common ancestor of all extant 143 cyanobacteria. This ancestor gave rise to basal lineages before diverging into macrocyanobacteria and 144 microcyanobacteria (Fig. 2). Although CuZnSODs are rare in macro-and micro-cyanobacteria, they are 145 present in most free-living basal lineages (Fig. 2). Two in particular (Pseudanabaena spp. and 146 Gloeobacter spp.) share sodC genes which are monophyletic (PP 1) and closely related in the same way 147 as the species are to one another (SI Appendix, Fig. S6). This suggests they have been vertically 148 inherited from the common ancestor of the cyanobacteria crown group.

149
As cyanobacteria diversified to occupy new ecological niches and habitats 35 , the sodC genes 150 which initially allowed crown cyanobacteria to use copper and zinc to protect against oxidative stress 151 were likely lost. Later HGTs likely occurred between non-cyanobacterial phyla and picocyanobacteria,

163
Estimates that include 95% credibility intervals allow for a range between 2.8 and 4.3 Ga, suggesting 164 that Cyanobacteria diverged from their sister phyla at the latest some 300 Myrs before the GOE (Table   165 2).

193
This conclusion is based on the close relationship of CuZnSODs from two basal lineages:

203
Timing the origin of Cu-based metalloenzymes is a difficult task. Characteristic protein-folds 204 required to bind Cu are predicted to have evolved during, or following, the GOE 29 . However, geological

232
The relatively late appearance of FeSOD and MnSOD in cyanobacteria is surprising as previous 233 studies have postulated an Archaean origin in bacteria 16,60,61 . What caused this delay? Earth system 234 models suggest that photoferrotrophs outcompeted cyanobacteria for upwelling nutrients in aquatic 235 habitats prior to the GOE 48 . Perhaps they also limited the soluble Fe 2+ available for cyanobacteria, thus 236 facilitating a selective advantage to lineages which used alternative metals for relieving oxidative stress.

237
As the atmosphere became more oxygenated, photoferrotrophs were marginalized to shrinking pools of 238 Fe 2+. Our analyses suggest that the Neoproterozoic oxygenation increased cyanobacteria's requirement 239 for ROS defense mechanisms so much that lineages began using FeSOD or MnSOD regardless of the 240 waning global concentrations of Fe and Mn (Fig. 3). Further study, however, will be needed to assess   Table 2. It is reassuring, however, that our estimates of all SOD isoforms predate the

294
Neoarchaean to Paleoproterozoic decline in Ni is also reflected in a compilation of shale data, normalized 295 to evolving upper continental crust (Fig. 4). Interestingly, however, there appears to be a slight uptick in 296 Ni in the terminal Neoproterozoic to Phanerozoic (Fig. 4) that occurs shortly after the appearance of 297 NiSOD in the Tonian period (Table 2)

344
Cyanobacterial SOD diversity was assessed by sampling 153 additional genomes (strains are 345 listed in SI Appendix, Table S1 and Table S5)

378
We also implemented microfossil calibrations as follows: filamentous cyanobacteria more than 1.

394
Chronology of SOD Isoforms

395
The order in which SOD isoforms appeared in cyanobacteria, was estimated using a method described 396 as 'topological comparison' 98 . First, a Bayesian protein phylogeny was created for each SOD isoform 397 using only cyanobacterial sequences (SI Appendix, Fig. S6 and Fig. S8). Bayesian phylogenetic 398 reconstructions of MnSODs and FeSODs had not converged after 2 weeks, so the sequences were S10). The resulting protein phylogenies were then compared to the species phylogeny of cyanobacteria 402 (Fig. 2) to identify monophyletic groups whose NiSOD, CuZnSOD or MnSOD/FeSOD had evolved as 403 expected by vertical inheritance (without HGT). The last common ancestor of each of these clades was 404 assumed to have been capable of using the corresponding SOD.

406
To find out which habitats ancestral cyanobacteria lived in when SOD isoforms appeared, Bayesian 407 stochastic character mapping 99 was implemented with SIMMAP v1.5 100 in the phytools package 101 of 408 R using our time-calibrated trees. Prior distribution on the root node of the tree was estimated based on 409 the data, 1000 simulations were performed and the all rates different model was utilised to allow transition 410 rates between marine and non-marine habitats to vary based the data (as implemented in 81 ). Character 411 states were coded as either 'marine' or 'non-marine' (SI Appendix, Table S1).

412
Compilation of Ni data         predicted to contain each SOD isoform is highlighted with a coloured circle and label (orange for NiSOD, 661 red for CuZnSOD, and blue for the ancestor of FeSOD and MnSOD; see Table 2 for posterior age 662 probabilities and SI Appendix, Fig. S11 for age distribution of these events