The Origin of Phototrophy Reveals the Importance of Pri- 1 ority Effects for Evolutionary Innovation

The history of life on Earth has been shaped by a series of major evolutionary innovations. 5 While some of these innovations occur repeatedly, some of the most important evolutionary 6 innovations (e.g., the origin of life itself, eukaryotes, or the genetic code) are evolutionary 7 singularities, arising just once in the history of life. This historical fact has often been in- 8 terpreted to mean that singularities are particularly difﬁcult, low-probability evolutionary 9 events, thus making the long-term course of life on Earth highly contingent on their chance 10 appearances. Alternatively, singularities may arise from evolutionary priority effects, where 11 ﬁrst-movers suppress subsequent independent origins. Here, we disentangle these hypothe- 12 ses by examining a distinctive innovation: phototrophy. The ability to use light to generate 13 metabolic energy evolved twice, preserving information about the origins of rare, transfor- 14 mative innovations that is lost when examining singular innovations. We show that the two 15 forms of phototrophy occupy opposite ends of several key trade-offs: efﬁciency of light cap- 16 ture vs. return on investment in protein infrastructure, dependence on limiting nutrients vs. 17 metabolic versatility, and complexity vs. simplicity. Our results suggest that the ‘dual singu- 18 larity’ of phototrophy exists due to evolutionary interactions between nascent phototrophs, 19 with phototrophic niche space too large for a ﬁrst mover to ﬁll all niches and fully suppress fu- 20

The origin of life, the freezing of the genetic code, and eukaryogenesis are evolutionary singularities as they were major innovations which occurred just once. Complex multicellularity has evolved at least six times across at least the last 1.6 billion years 2, 3, 5 . Phototrophic metabolism has evolved twice, via chlorophototrophy and retinalophototrophy. Chlorophototrophy dates to at least 3.5 billion years ago with the oldest unequivocal photosynthetic microbial mats 16, 17 , though some argue for older dates 18 . The origin of retinalophototrophy is uncertain due to its lack of preservation in the fossil record, and could date from anywhere between the Hadean to shortly before the rise of animals, but is more likely to be ancient 1920 . This dual singularity provides unique insight into the nature and process of evolutionary innovation. between these mechanisms. In this paper, we circumvent these limitations by examining evolution on the potential primary production of biomass in a nonphotosynthetic biosphere (Supplemental 69 Figure 1). Photosynthesis is thus the key factor allowing the existence of the large, high-biomass, 70 geochemically significant modern biosphere, transforming the composition of both the atmosphere 24 71 and the geosphere 25 over geological time. 72 Unlike other biosphere-transforming innovations, the ability to use light for metabolic energy 73 appears to have evolved independently twice, via retinalophototrophy and chlorophototrophy. As 74 the only such 'dual singularity', it preserves information on the evolutionary factors underpinning 75 the origin of rare, impactful innovations that have been lost in true singularities. By examining their 76 properties and evolutionary histories, we find that chlorophototrophy and retinalophototrophy have 77 precisely partitioned phototrophic niche space. They occupy opposite ends of critical trade-offs 78 between efficiency per unit resource versus efficiency per unit infrastructure, use of rare limiting 79 nutrients versus metabolic versatility, and complexity versus simplicity. This deep complementarity 80 suggests that phototrophy has evolved twice because phototrophic niche space is too large for 81 an initial first mover to fully suppress future innovation, but too small to support many separate 82 innovations. Together, this work highlights the critical role of evolutionary priority effects in the 83 evolution of biological innovations, and suggests that the origins of evolutionary singularities may 84 be less constrained or contingent than is widely believed. : Simplified illustration of the three main types of phototrophic metabolism. A) Chlorophototrophy, type I reaction center. A photon is absorbed by one of a diverse array of antenna complexes (dark green) and passed as an exciton via Förster resonance to chlorophyll or bacteriochlorophyll molecules (small dark green spots) within the dimeric photosynthetic reaction center (light green). A type I reaction center is illustrated acquiring an electron via a cytochrome derived from an environmental reducing agent, boosting it via light energy to a low redox potential, and passing it via iron-sulfur clusters (red) to ferredoxin (brown) which can be used to build biomass via carbon and nitrogen fixation. B) Chlorophototrophy, type II reaction center illustrated passing electrons to a quinone electron acceptor, allowing for simple cyclic electron transfer via cytochrome bc1 (complex III) (orange) and the pumping of two protons per absorbed photon. C) Retinalophototrophy. A single molecule of retinal (purple) is bound to a microbial rhodopsin membrane protein (pink). Absorption of a photon causes one proton to be pumped the exterior of the cell, upon which it can participate in chemiosmotic ATP production via the membrane ATP synthase (blue). Some rhodopsins, not directly involved in phototrophy, are also capable of pumping ions such as 144 chloride or sodium and others function as light sensors 48 . While there are no known autotrophs 145 able to fix biomass from CO 2 using only the energy derived from microbial rhodopsins, the energy 146 generated by this system appears to be quite important for many photoheterotrophs. This energy 147 can prevent starvation in marine bacteria 49, 50 , and is extensively used to supplement heterotrophic 148 metabolism: the quantity of light absorbed by retinalophototrophs in the ocean is thought to be at 149 least as large as that absorbed by chlorophototrophs 51 .

150
The phylogenetic ubiquity of microbial rhodopsins, in contrast to the patchy distribution of 151 chlorophototrophy, has only been fully appreciated in the last two decades. Approximately half 152 of marine bacterial cells, from many taxa, bear diverse bacterial rhodopsin genes 52, 53 . In addition 153 to haloarchaea, they are present in marine bacteria 54 , marine archaea 55 , fungi 56 , and heterotrophic 154 marine eukaryotes 57-59 . They are known to acidify cellular compartments via pumping protons, 155 and in some taxa are among the most highly expressed proteins 59 , contributing significantly to the 156 cell's energy budget. Rhodopsins have even been discovered in metagenomes of Heimdallarchaea, a member of the Asgard archaea considered a likely sister to the archaeal ancestor of eukaryotes 60 , 158 and in numerous marine viruses 61, 62 .  (Table 1). By abstracting away from their fine details and looking at 172 gross compositions and the products of their metabolisms ( Figure 3) the major differences between 173 them may be understood more easily.

174
One of the greatest differences between retinalophototrophs and chlorophototrophs is the effi-  Comparison of ecologically and evolutionarily relevant differences between chlorophototrophic reaction-center-based and retinalophototrophic microbial-rhodopsin-based phototrophy. in protein mass -the infrastructure must be built before it can transduce energy or nutrients, and 253 has a finite lifetime before being either recycled or diluted away by growth and division. As such, 254 every metabolic pathway also has a rate of return on investment. In order to determine the return on 255 investment available from multiple phototrophic pathways, one must take into account the mass per 256 functional unit, the rate of operation of the protein machinery, and the yield per cycle, yielding a 257 specific energy flux per unit protein mass.
258 Figure 4: Ecological comparison between chlorophototrophy and retinalophototrophy. A) Calculated maximum pumped proton flux available per kDa of protein mass of anoxygenic purple bacterial reaction centers, oxygenic reaction centers, proteorhodopsin, and bacteriorhodopsin at saturating levels of light. Microbial rhodopsins saturate at much higher specific metabolic energy fluxes than reaction centers. B) Calculated proton flux available per kilodalton of mass of different phototrophic systems at different light intensities. Chlorophototrophic reaction centers produce more energy flux at low light levels compared to microbial rhodopsins, but saturate quickly, while microbial rhodopsins function best at high light levels with higher specific metabolic energy fluxes. C) Calculated maximum pumped proton flux available per kDa per unit incident light in microeinsteins per square meter. Chlorophototrophic reaction centers are capable of extracting much more energy flux per unit incident light. D) Calculated energy flux per kilodalton of machinery per microeinstein of incident light. Chlorophototrophic reaction centers are significantly more efficient per unit incident light when light is scarce, but rapidly saturate due to having large absorption cross sections per reaction center, thereby gathering more light than can be converted to energy. Microbial rhodopsins, on the other hand, are significantly less efficient per photon, but use light more efficiently than chlorophototrophy when light levels are high. See Supplement S1 for calculations. We quantified the effective flux in terms of protons pumped per kilodalton per second at saturating 262 light levels (See Table 2 and Supplement S1). Despite their higher efficiency per photon absorbed 263 and faster photocycle, chlorophototrophic machinery is so much more massive than microbial 264 rhodopsins that their specific energy flux per unit mass is significantly lower. Proteorhodopsin and 265 bacteriorhodopsin are calculated as 2.78-fold and 5.76-fold more efficient per unit investment than 266 oxygenic RCs respectively, with anoxygenic RCs roughly equivalent to oxygenic RCs (Table 2 267 Figure 4 A, and Supplement S1).

285
While chlorophototrophic reaction centers are more efficient per unit ambient light in the limit of 286 low light, this is reversed at higher light levels.

287
The differences between chlorophototrophy and retinalophototrophy stem from an intrinsic bio-288 physical trade-off. It is not possible to build a phototrophic system that has both high metabolic is seen more frequently than the ED pathway in obligate anaerobes which cannot switch to aerobic 311 respiration and must obtain more energy from their limited available substrate.  Figure 5: Schematic illustration of hypothesized evolutionary history of phototrophic metabolism on Earth. All modern chlorophototrophs (dark green circles) and retinalophototrophs (dark purple diamonds) lay roughly along a 'Pareto front' representing a trade-off between energy captured per incident photon and wattage available per unit phototrophic infrastructure. Representative differences are illustrated above the curve, with differences in antenna pigments (green and purple funnels) and electron transport chains (orange oval accessory components) contributing to differences in energy flux and energy yield within each class of phototrophs. Early chlorophototrophs (light green circles) and retinalophototrophs (light purple diamonds) lay far away from this Pareto front, and rapidly evolved towards it, subsequently diversifying along the front (arrows). An architectural limitation (red line) prevented whichever evolved first from diversifying to fill all positions on the trade-off curve Pareto front, allowing a second novel phototroph sufficiently different to evolve and fill the rest of the Pareto front. Each phototrophic metabolism suppresses the evolution of novel unrelated phototrophic pathways that are ecologically similar to it but strictly inferior in their initial, unoptimized forms (red Xs).
restructuring. Without any redox-active cofactors in its structure, it cannot be recruited to interact 352 with electron transport chains or redox metabolism. Rhodopsin thus appears to be incapable of 353 evolving to pump more than one proton per photon and efficiently using available light resources, 354 although its small mass means it enjoys a high maximum energy flux per unit mass.

355
Conversely, the mass of the core machinery of the chlorophototrophic reaction center appears 356 to be constrained, such that it cannot be reduced below a relatively large minimum size. While