The oxygen-transporting protein haemo- globin has undergone repeated adaptations as animals evolved to conquer new environ- ments — from the depths of the oceans 1 to high mountain ranges 2 . These adaptations relied on changes in the long-range interactions between oxygen-binding sites buried in the protein’s subunits, and between these regions and binding sites for a multitude of small effec- tor molecules on the protein’s surface 3 . How did this complex molecular machine, which can respond so exquisitely to available levels of both oxygen and several other effector molecules, come into being? On page 480, Pillai et al. 4 reconstruct the stepwise evolution of haemoglobin from precursors that existed more than 400 million years ago. Almost nothing was previously known about how the four-subunit (tetrameric) form of haemoglobin that is found in modern-day jawed vertebrates evolved from ancient mono- mers. Tetrameric haemoglobin consists of two α- and two β-subunits. Pillai et al. computa- tionally reconstructed an evolutionary tree to chart the protein’s ancient history, using the amino-acid sequences of a large collection of the closely related vertebrate globin proteins, which exist as either monomers or tetramers. Molecular biology Evolution of a molecular machine Michael Berenbrink The multi-subunit protein haemoglobin relies on complex interactions between its components to function properly. Analysis of ancient precursors suggests that its evolution from a simple monomer involved only a few steps. See p.480 The authors’ tree was constructed taking into account that amino-acid substitutions a given protein shares with close relatives tend to have originated in more-recent common ancestors than have those it shares with more-distant relatives. The reconstructed evolutionary tree indicates that multiple rounds of gene duplication and subsequent divergence gave rise to the globin family and, by way of several ancestral proteins, to tetrameric haemoglobin (Fig. 1). What is special about the study is that Pillai and colleagues went on to resurrect several of these extinct ancestral proteins, generating them from the amino-acid sequences pre- dicted by the tree. The group then tested these proteins’ functions. First, Pillai and colleagues analysed whether each ancestral protein could form dimers and tetramers of like or unlike subunits. The earliest protein — a common ancestor of haemoglobin and the monomeric globin pro- tein myoglobin, named AncMH by the authors — exists only as a monomer. A later protein, named Ancα/β, which is the ancestor of all existing haemoglobin subunits, forms homo- dimers when expressed at high levels. The authors’ tree indicates that Ancα/β underwent gene duplication to produce two proteins: the ancestors of all existing α- or β-subunits, which the group respectively named Ancα and Ancβ. These proteins also form homodimers, or even homotetramers, when expressed alone. However, when the two are expressed together in equal proportions, they can form heterodimers, which then further align to yield haemoglobin tetramers. The group next investigated the oxygen- binding affinity of the ancestral proteins, along with their oxygen cooperativity (the ability of oxygen-binding subunits to interact with one another) and their ‘allosteric’ regula- tion by a potent, artificial effector molecule, inositol hexaphosphate (IHP). They found that only Ancα and Ancβ — when expressed together at high concentrations — show simi- lar oxygen-binding affinity, cooperativity and allosteric regulation to today’s haemoglobin protein. These features are shared by all living jawed vertebrates, but are absent or achieved in a different way in jawless vertebrates, whose haemoglobin proteins are of more ancient ori- gin. This indicates that the basic functions of jawed-vertebrate haemoglobin had already evolved in a common ancestor of these animals but at some time after the split with jawless vertebrates. Next, Pillai et al. modelled the stepwise changes in α- and β-subunit interfaces that might have allowed Ancα and Ancβ first to form heterodimers with one another, and later heterotetramers from pairs of such dimers. The modelling indicated that strikingly few amino-acid substitutions might have been needed to transform a simple monomeric for the hydrogen-producing reaction in a protective shell of a chromium compound. This combination of complex mitigation strategies proved highly successful: the authors reported EQEs of up to 96% when their photocatalysts were irradiated with light in the wavelength range of 350–360 nanometres. This is excellent news, because it means they have designed an almost perfect photocatalyst — the IQE must be between 96% and 100%. This is a spectacular result for several reasons, even though strontium titanate is ‘just’ a model system for visible-light photocatalysts. First, it demonstrates that experiments can be designed in which EQEs come close to IQEs within an acceptable error margin of less than 4%. Improved experimental set-ups in which measured EQEs are very near to IQEs should facilitate the comparison of photocatalysts and therefore accelerate progress in this field. Second, it proves that the combination of design strategies used by the authors can indeed eliminate efficiency losses associated with recombination. It is to be expected that the strategies used to improve the efficiency of strontium titanate will also apply to photo- catalysts driven by visible light — and could therefore enable the conversion of solar energy to hydrogen with efficiencies of about 10%. Finally, and most importantly, Takata and colleagues’ findings will inspire and encourage other researchers to continue their work on photocatalysts. One of the authors of the work, Kazunari Domen, published his first paper 9 on the use of strontium titanate as a photocatalyst in 1980. This shows the timescale needed for success in this area. Although we do not yet have a route for the sustainable and econom- ically viable production of hydrogen, we stand a good chance of finding one in the next few decades. This paper vouches for it. Simone Pokrant is in the Department of Chemistry and Physics of Materials, University of Salzburg, Salzburg A-5020, Austria. e-mail: simone.pokrant@sbg.ac.at 1. Takata, T. et al. Nature 581, 411–414 (2020). 2. Abbott, D. Proc. IEEE 98, 42–66 (2010). 3. Ochoa Robles, J., De-León Almaraz, S. & Azzaro-Pantel, C. in Hydrogen Supply Chains 8–11 (Academic, 2018). 4. Pinaud, B. A. et al. Energy Environ. Sci. 6, 1983–2002 (2013). 5. Wrighton, M. S., Wolczanski, P. T. & Ellis, A. B. J. Solid State Chem. 22, 17–29 (1977). 6. Ham, Y. et al. J. Mater. Chem. A 4, 3027–3033 (2016). 7. Giocondi, J. L. & Rohrer, G. S. J. Am. Ceram. Soc. 86, 1182–1189 (2003). 8. Mu, L. et al. Energy Environ. Sci. 9, 2463–2469 (2016). 9. Domen, K., Naito, S., Soma, M., Onishi, T. & Tamaru, K. J. Chem. Soc. Chem. Commun. 543–544 (1980). “Pillai and colleagues’ work serves as one of the clearest examples so far of how such complexity can arise.” 388 | Nature | Vol 581 | 28 May 2020 News & views ©2020SpringerNatureLimited.Allrightsreserved.