《生物技术与人类》课程教学资源（经典阅读）Evolution of increased complexity in a molecular machine

LETTER doi:10.1038/nature10724 Evolution of increased complexity in a molecular machine Gregory C.Finnigan!*,Victor Hanson-Smith2.3+,Tom H.Stevens!&Joseph W.Thornton2.4.5 Many cellular processes are carried out by molecular'machines'- Comparative genomic approaches suggest that the components of assemblies of multiple differentiated proteins that physically inter- many molecular machines have appeared sequentially during evolu- act to execute biological functions'-.Despite much speculation, tion and that complexity increased gradually by incorporating new strong evidence of the mechanisms by which these assemblies parts into simpler assemblies-Such horizontal analyses of extant evolved is lacking.Here we use ancestral gene resurrection and systems,however,cannot decisively test these hypotheses or reveal manipulative genetic experiments to determine how the complexity the mechanisms by which additional parts became obligate com- of an essential molecular machine-the hexameric transmembrane ponents of larger complexes.In contrast,vertical approaches that ring of the eukaryotic V-ATPase proton pump-increased hundreds combine computational phylogenetic analysis with gene synthesis of millions of years ago.We show that the ring of Fungi,which is and molecular assays allow changes in the sequence,structure and composed of three paralogous proteins,evolved from a more ancient function of reconstructed ancestral proteins to be experimentally two-paralogue complex because of a gene duplication that was traced through time Here we apply this approach to characterize followed by loss in each daughter copy of specific interfaces by the evolution of a small molecular machine and dissect the mechan- which it interacts with other ring proteins.These losses were com- isms that caused it to increase in complexity. plementary,so both copies became obligate components with The vacuolar H-ATPase (V-ATPase)is a multisubunit protein restricted spatial roles in the complex.Reintroducing a single complex that pumps protons across membranes to acidify subcellular historical mutation from each paralogue lineage into the resurrected compartments;this function is required for intracellular protein traf- ancestral proteins is sufficient to recapitulate their asymmetric ficking,coupled transport of small molecules and receptor-mediated degeneration and trigger the requirement for the more elaborate endocytosis'.V-ATPase dysfunction has been implicated in human three-component ring.Our experiments show that increased com- osteoporosis,in acquired drug resistance in human tumours,and in plexity in an essential molecular machine evolved because of simple, pathogen virulence A key subcomplex of the V-ATPase is the Vo high-probability evolutionary processes,without the apparent protein ring,a hexameric assembly that uses a rotary mechanism to evolution of novel functions.They point to a plausible mechanism move protons across organelle membranes(Fig.1a)6.Although the for the evolution of complexity in other multi-paralogue protein V-ATPase is found in all eukaryotes,the Vo ring varies in subunit complexes. composition among lineages.In animals and most other eukaryotes, Amoebozoa, Apicomplexa Animals, Choanoflagellates Subunit Anc 6 Fungi Amoebozoa -Apicomplexa Animals. Choanoflagellates brane Subunit Anc.3-11. Fungi Anc.3 Subunit 0.8 subs/site ● H+ 11 Anc.11 Figure 1 Structure and evolution of the V-ATPase complex.a,In S. contain subunit 11.Circles show ancestral proteins reconstructed in this study. cerevisiae,the V-ATPase contains two subcomplexes:the octameric V domain is Colours correspond to those of subunits in panel a;unduplicated orthologues of on the cytosolic side of the organelle membrane,and the hexameric Vo ring is Vma3 and Vmall are green.Asterisks show approximate likelihood ratios for membrane bound.Protein subunits Vma3,Vmall and Vmal6 are labelled and major nodes*来，>103，*米，>10：*米，>10来，<10，~，<2.The complete coloured.b,Maximum likelihood phylogeny of V-ATPase subunits Vma3, phylogeny is presented in Supplementary Information,section 2. Vmall and Vma16.All eukaryotes contain subunits 3 and 16,but Fungi also Institute of Molecular Biology,University of Oregon,Eugene,Oregon 97403,USA2Institute for Ecology and Evolution,University of Oregon,Eugene.Oregon 97403.USADepartment of Computer and Information Science,University of Oregon.Eugene.Oregon 97403.USA Howard Hughes Medical Institute.Eugene,Oregon 97403,USASDepartments of Human Genetics and Ecology&Evolution. University of Chicago,Chicago,llinois 60637.USA These authors contributed equally to this work. 360 NATURE I VOL 48119 JANUARY 2012 2012 Macmillan Publishers Limited.All rights reserved

LETTER doi:10.1038/nature10724 Evolution of increased complexity in a molecular machine Gregory C. Finnigan1 *, Victor Hanson-Smith2,3*, Tom H. Stevens1 & Joseph W. Thornton2,4,5 Many cellular processes are carried out by molecular ‘machines’— assemblies of multiple differentiated proteins that physically inter￾act to execute biological functions1–8. Despite much speculation, strong evidence of the mechanisms by which these assemblies evolved is lacking. Here we use ancestral gene resurrection9–11 and manipulative genetic experiments to determine how the complexity of an essential molecular machine—the hexameric transmembrane ring of the eukaryotic V-ATPase proton pump—increased hundreds of millions of years ago. We show that the ring of Fungi, which is composed of three paralogous proteins, evolved from a more ancient two-paralogue complex because of a gene duplication that was followed by loss in each daughter copy of specific interfaces by which it interacts with other ring proteins. These losses were com￾plementary, so both copies became obligate components with restricted spatial roles in the complex. Reintroducing a single historical mutation from each paralogue lineage into the resurrected ancestral proteins is sufficient to recapitulate their asymmetric degeneration and trigger the requirement for the more elaborate three-component ring. Our experiments show that increased com￾plexity in an essential molecular machine evolved because of simple, high-probability evolutionary processes, without the apparent evolution of novel functions. They point to a plausible mechanism for the evolution of complexity in other multi-paralogue protein complexes. Comparative genomic approaches suggest that the components of many molecular machines have appeared sequentially during evolu￾tion and that complexity increased gradually by incorporating new parts into simpler assemblies2–8. Such horizontal analyses of extant systems, however, cannot decisively test these hypotheses or reveal the mechanisms by which additional parts became obligate com￾ponents of larger complexes. In contrast, vertical approaches that combine computational phylogenetic analysis with gene synthesis and molecular assays allow changes in the sequence, structure and function of reconstructed ancestral proteins to be experimentally traced through time.9–11 Here we apply this approach to characterize the evolution of a small molecular machine and dissect the mechan￾isms that caused it to increase in complexity. The vacuolar H1-ATPase (V-ATPase) is a multisubunit protein complex that pumps protons across membranes to acidify subcellular compartments; this function is required for intracellular protein traf￾ficking, coupled transport of small molecules and receptor-mediated endocytosis1 . V-ATPase dysfunction has been implicated in human osteoporosis, in acquired drug resistance in human tumours, and in pathogen virulence12–14. A key subcomplex of the V-ATPase is the V0 protein ring, a hexameric assembly that uses a rotary mechanism to move protons across organelle membranes (Fig. 1a)15,16. Although the V-ATPase is found in all eukaryotes, the V0 ring varies in subunit composition among lineages. In animals and most other eukaryotes, 1 Institute of Molecular Biology, University of Oregon, Eugene, Oregon 97403, USA. 2 Institute for Ecology and Evolution, University of Oregon, Eugene, Oregon 97403, USA. 3 Department of Computer and Information Science, University of Oregon, Eugene, Oregon 97403, USA. 4 Howard Hughes Medical Institute, Eugene, Oregon 97403, USA. 5 Departments of Human Genetics and Ecology & Evolution, University of Chicago, Chicago, Illinois 60637, USA. *These authors contributed equally to this work. Membrane **** V1 H a 11 16 3 Animals, Choanoflagellates Fungi 0.8 subs/site Fungi Anc.3-11 Amoebozoa, Apicomplexa Anc.11 Anc.3 ** **** **** ** Amoebozoa, Apicomplexa Fungi Anc.16 Animals, Choanoflagellates **** *** **** b V0 + **** **** **** ~ *** * * * * Subunit 3 Subunit 11 Subunit 16 Figure 1 | Structure and evolution of the V-ATPase complex. a, In S. cerevisiae, the V-ATPase contains two subcomplexes: the octameric V1 domain is on the cytosolic side of the organelle membrane, and the hexameric V0 ring is membrane bound. Protein subunits Vma3, Vma11 and Vma16 are labelled and coloured. b, Maximum likelihood phylogeny of V-ATPase subunits Vma3, Vma11 and Vma16. All eukaryotes contain subunits 3 and 16, but Fungi also contain subunit 11. Circles show ancestral proteins reconstructed in this study. Colours correspond to those of subunits in panel a; unduplicated orthologues of Vma3 and Vma11 are green. Asterisks show approximate likelihood ratios for major nodes: ****, .103 ; ***, .102 ; **, .10; *, ,10; ,, ,2. The complete phylogeny is presented in Supplementary Information, section 2. 360 | NATURE | VOL 481 | 19 JANUARY 2012 ©2012 Macmillan Publishers Limited. All rights reserved

ETTER RESEARCH the ring consists of one subunit of Vmal6 protein and five copies of its the presence of elevated CaCl,was rescued,indicating that the func- paralogue,Vma3(Fig.1b)'.In Fungi,the ring consists of one Vma16 tions of the present-day Vma3 and Vmall proteins were already pre- subunit,four copies of Vma3 and one Vmall subunit,arranged in a sent before the duplication that generated them(Fig.2a).Furthermore, specific orientation.All three proteins are required for V-ATPase Anc.3-11-unlike either of its present-day descendants-can partially function in Fungi1s,but the mechanisms are unknown by which rescue growth in yeast that are doubly deficient for both Vma3 and both Vma3 and Vmall became obligate components with specific Vmall (vma34 vmall4).The reconstructed Anc.16 also rescued positional roles in the complex. growth in Vmal6-deficient S.cerevisiae (vmal64)(Fig.2b),and co- To understand how the three-component ring evolved,we recon- expression of Anc.3-11 and Anc.16 together rescued cell growth in structed ancestral Vo proteins from just before and after the increase in vma34 vmal14 vma164 yeast,which lack all three ring subunits complexity,synthesized and functionally characterized them in a yeast (Fig.2c).The ancestral genes specifically restore proper V-ATPase genetic system,and used manipulative methods to identify the genetic function in acidification of the vacuolar lumen (Fig.2g).In addition, and molecular mechanisms by which their functions changed.We first mutation of the ancestral subunits to remove glutamic acid residues inferred the phylogeny and best-fit evolutionary model of the protein known to be essential for V-ATPase enzyme function'323 abolished family of which Vma3,Vmall and Vmal6 are members,using the their ability to rescue growth on CaCl2(Supplementary Information, sequences ofall 139 extant family members available in GenBank(Sup- section 7).These inferences about the functions of Anc.3-11 and Anc.16 plementary Table 1).The maximum likelihood phylogeny(Fig.1band are robust to uncertainty about ancestral amino acid states.We recon- Supplementary Information,section 2)indicates that Vma3 and structed alternative versions of Anc.3-11 and Anc.16 by introducing Vmall are sister proteins that were produced by duplication of an amino acid states with posterior probability >0.2,but none of these ancestral gene (Anc.3-11)before the last common ancestor of all abolished the ability of the ancestral genes to substitute functionally for Fungi(~800 million years ago2).Whether this duplication occurred the extant subunits (Supplementary Information,section 8).These before or after the divergence of Fungi from other eukaryotes (~1 results establish that during the increase in complexity,neither the Vo billion years ago)is not clearly resolved,although the latter scenario complex nor its component proteins evolved new functions required is more parsimonious.The Vma3/Vmall and Vmal6 lineages,in turn, for growth under the conditions in which the ring is known to be descend from an older gene duplication deep in the eukaryotic lineage important. (Fig.Ib).We used a maximum likelihood algorithm2 to infer the Similar experiments with the components of the ancestral three- ancestral amino acid sequences with the highest probability of pro- component ring show that after the duplication of Anc.3-11,its ducing all the extant sequence data,given the best-fit phylogeny and descendants Anc.3 and Anc.11 both became necessary for a functional model.We reconstructed the ancestral proteins(Anc.3-11 and Anc.16) complex because of complementary losses of ancestral functions that made up the ancient two-paralogue eukaryotic ring,as well as the Unlike Anc.3-11,expression of Anc.3 can rescue growth and vacuole duplicated subunits Anc.3 and Anc.11 from the three-component ring acidification in vma34 but not vmal14 yeast,and Anc.11 can rescue in the common ancestor of all Fungi (Supplementary Information, growth in vmal14 but not vma34 yeast (Fig.2d,e,g).Furthermore, sections 3 and 4). both Anc.3 and Anc.11 are required to rescue growth fully in To characterize the functions of these reconstructed proteins,we syn- vma34 vmal14 yeast (Fig.2f).These data indicate that after its origin thesized coding sequences and transformed them into Saccharomyces by gene duplication,Anc.11 lost the ancestral protein's ability to carry cerevisiae deficient for various ring components and therefore incapable out at least some functions of Vma3,and Anc.3 lost the ancestral of growth in the presence ofelevated CaCl,(ref.22).We found that the capacity to carry out those of Vmall. ancestral two-subunit ring can functionally replace the three-subunit We conjectured that Vma3 and Vmall evolved their specialized ring of extant yeast.When the resurrected Anc.3-11 was transformed roles because they lost specific interfaces present in their ancestor that into yeast deficient for Vma3 (vma34)or Vmall(vmal14),growth in are required for ring assembly.Previous experiments with fusions of Plasmid Genotype YEPD CaCL, Plasmid Genotype YEPD CaCl2 9 Quinacrine DIC None ●● None WT ● None 34 None 34 None Anc.3-11 Anc.3-11 ●金 ● ● Anc.11 34 Anc.3-11 114 ●●● Anc.3-11 11d Anc.3-11 34114 ● ●●● Anc.11 11 Anc.3-11 Anc.3-11 b None WT None 164 Anc.3 Anc.16 164 Anc.3-11 114 Anc.3 None Anc.3 3411A None 34114164 Anc.11 3411A Anc.3-11/Anc.16 34114164 Anc.3/Anc.11 34114 ●● Figure 2 Two reconstructed ancestral Vo subunits functionally replace the Vma3,Vmall and Vmal6.d,Anc.11 rescues growth in vmal14 but not in three-paralogue ring in extant yeast.S.cerevisiae were plated in decreasing vma34 yeast.e,Anc.3 rescues growth in vma34 but not vmal14 yeast.f,Anc.3 concentrations on permissive medium (YEPD)buffered with elevated CaCl2. and Anc.11 together rescue growth in vma34 vmal14 mutants. a,Expression of Anc.3-11 rescues growth in yeast that are deficient for g,Yeast expressing reconstructed ancestral subunits properly acidified the endogenous subunit Vma3 (34),subunit Vmall (114)or both (34 114). vacuolar lumen.Red signal shows yeast cell walls;green signal (quinacrine) Growth of wild-type (WT)yeast is shown for comparison.b,Anc.16 rescues shows acidified compartments.Yeast were visualized by differential growth in yeast that are deficient for subunit Vmal6(164).c,Expression of interference contrast microscopy Anc.3-11 and Anc.16 together rescues growth in yeast that are deficient for 19 JANUARY 2012 VOL 481 NATURE I 361 2012 Macmillan Publishers Limited.All rights reserved

the ring consists of one subunit of Vma16 protein and five copies of its paralogue, Vma3 (Fig. 1b)1 . In Fungi, the ring consists of one Vma16 subunit, four copies of Vma3 and one Vma11 subunit, arranged in a specific orientation17. All three proteins are required for V-ATPase function in Fungi18,19, but the mechanisms are unknown by which both Vma3 and Vma11 became obligate components with specific positional roles in the complex. To understand how the three-component ring evolved, we recon￾structed ancestral V0 proteins from just before and after the increase in complexity, synthesized and functionally characterized them in a yeast genetic system, and used manipulative methods to identify the genetic and molecular mechanisms by which their functions changed. We first inferred the phylogeny and best-fit evolutionary model of the protein family of which Vma3, Vma11 and Vma16 are members, using the sequences of all 139 extant family members available in GenBank (Sup￾plementary Table 1). The maximum likelihood phylogeny (Fig. 1b and Supplementary Information, section 2) indicates that Vma3 and Vma11 are sister proteins that were produced by duplication of an ancestral gene (Anc.3-11) before the last common ancestor of all Fungi (,800 million years ago20). Whether this duplication occurred before or after the divergence of Fungi from other eukaryotes (,1 billion years ago20) is not clearly resolved, although the latter scenario is more parsimonious. The Vma3/Vma11 and Vma16 lineages, in turn, descend from an older gene duplication deep in the eukaryotic lineage (Fig. 1b). We used a maximum likelihood algorithm21 to infer the ancestral amino acid sequences with the highest probability of pro￾ducing all the extant sequence data, given the best-fit phylogeny and model.We reconstructed the ancestral proteins (Anc.3-11 and Anc.16) that made up the ancient two-paralogue eukaryotic ring, as well as the duplicated subunits Anc.3 and Anc.11 from the three-component ring in the common ancestor of all Fungi (Supplementary Information, sections 3 and 4). To characterize the functions of these reconstructed proteins, we syn￾thesized coding sequences and transformed them into Saccharomyces cerevisiae deficientfor various ring components and therefore incapable of growth in the presence of elevated CaCl2 (ref. 22). We found that the ancestral two-subunit ring can functionally replace the three-subunit ring of extant yeast. When the resurrected Anc.3-11 was transformed into yeast deficient for Vma3 (vma3D) or Vma11(vma11D), growth in the presence of elevated CaCl2 was rescued, indicating that the func￾tions of the present-day Vma3 and Vma11 proteins were already pre￾sent before the duplication that generated them (Fig. 2a). Furthermore, Anc.3-11—unlike either of its present-day descendants—can partially rescue growth in yeast that are doubly deficient for both Vma3 and Vma11 (vma3Dvma11D). The reconstructed Anc.16 also rescued growth in Vma16-deficient S. cerevisiae (vma16D) (Fig. 2b), and co￾expression of Anc.3-11 and Anc.16 together rescued cell growth in vma3Dvma11Dvma16D yeast, which lack all three ring subunits (Fig. 2c). The ancestral genes specifically restore proper V-ATPase function in acidification of the vacuolar lumen (Fig. 2g). In addition, mutation of the ancestral subunits to remove glutamic acid residues known to be essential for V-ATPase enzyme function17,23 abolished their ability to rescue growth on CaCl2 (Supplementary Information, section 7). These inferences about the functions of Anc.3-11 and Anc.16 are robust to uncertainty about ancestral amino acid states. We recon￾structed alternative versions of Anc.3-11 and Anc.16 by introducing amino acid states with posterior probability .0.2, but none of these abolished the ability of the ancestral genes to substitute functionally for the extant subunits (Supplementary Information, section 8). These results establish that during the increase in complexity, neither the V0 complex nor its component proteins evolved new functions required for growth under the conditions in which the ring is known to be important. Similar experiments with the components of the ancestral three￾component ring show that after the duplication of Anc.3-11, its descendants Anc.3 and Anc.11 both became necessary for a functional complex because of complementary losses of ancestral functions. Unlike Anc.3-11, expression of Anc.3 can rescue growth and vacuole acidification in vma3D but not vma11D yeast, and Anc.11 can rescue growth in vma11D but not vma3D yeast (Fig. 2d, e, g). Furthermore, both Anc.3 and Anc.11 are required to rescue growth fully in vma3Dvma11D yeast (Fig. 2f). These data indicate that after its origin by gene duplication, Anc.11 lost the ancestral protein’s ability to carry out at least some functions of Vma3, and Anc.3 lost the ancestral capacity to carry out those of Vma11. We conjectured that Vma3 and Vma11 evolved their specialized roles because they lost specific interfaces present in their ancestor that are required for ring assembly. Previous experiments with fusions of WT Quinacrine DIC Anc.3-11 Anc.11 WT Anc.3-11 Anc.11 Genotype YEPD CaCl2 Anc.3-11 Anc.3 Anc.3-11 Anc.3 Plasmid Anc.3 Anc.11 YEPD CaCl2 WT Plasmid Anc.3-11 Anc.3-11 Anc.3-11 a c WT WT Anc.16 b 11Δ 11Δ 3Δ 3Δ Anc.3-11/Anc.16 Anc.3/Anc.11 Genotype 3Δ 11Δ 16Δ 16Δ 3Δ 11Δ 16Δ 3Δ 3Δ 3Δ 3Δ 3Δ 11Δ 11Δ 3Δ 11Δ 3Δ 11Δ 3Δ 11Δ 11Δ 11Δ d e f g 3Δ 3Δ + Anc.3-11 16Δ + Anc.16 11Δ + Anc.11 3Δ + Anc.3 None None None None None None None None None 3Δ 11Δ 16Δ Figure 2 | Two reconstructed ancestral V0 subunits functionally replace the three-paralogue ring in extant yeast. S. cerevisiae were plated in decreasing concentrations on permissive medium (YEPD) buffered with elevated CaCl2. a, Expression of Anc.3-11 rescues growth in yeast that are deficient for endogenous subunit Vma3 (3D), subunit Vma11 (11D) or both (3D11D). Growth of wild-type (WT) yeast is shown for comparison. b, Anc.16 rescues growth in yeast that are deficient for subunit Vma16 (16D). c, Expression of Anc.3-11 and Anc.16 together rescues growth in yeast that are deficient for Vma3, Vma11 and Vma16. d, Anc.11 rescues growth in vma11D but not in vma3D yeast. e, Anc.3 rescues growth in vma3D but not vma11D yeast. f, Anc.3 and Anc.11 together rescue growth in vma3Dvma11D mutants. g,Yeast expressing reconstructed ancestral subunits properly acidified the vacuolar lumen. Red signal shows yeast cell walls; green signal (quinacrine) shows acidified compartments. Yeast were visualized by differential interference contrast microscopy. LETTER RESEARCH 19 JANUARY 2012 | VOL 481 | NATURE | 361 ©2012 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER extant yeast proteins have shown that the arrangement of subunits in Vmal6(Fig.3e).Anc.11 functioned when constrained to participate in the ring is constrained by the capacity of each subunit to form specific interface R with Vmal6 and interface P with Vma3,but ring function interfaces (which we labelled P,Q and R)with the other subunits24. was lost when Anc.11 was constrained to participate in interface Q Specifically,Vmall is restricted to a single position between Vmal6 with Vmal6 and interface P with Vma3.This result indicates that and Vma3,because its clockwise interface can participate only in Anc.11 lost the capacity to form one or both of these interfaces during interface R with Vmal6,and its anticlockswise interface can participate its post-duplication divergence from Anc.3-11(Fig.3f). only in interface P with the clockwise side of Vma3(Fig.3).By contrast, Taken together,these data indicate that the specificity of the ring copies of Vma3 occupy several positions in the ring,because they form arrangement and the obligate roles of Vma3 and Vmall evolved by interface P with other copies of Vma3 or Vmall,as well as interface Q complementary loss of asymmetric interactions with other members with Vmal6.However,Vma3 cannot form interface R with Vmal6.As of the ring (Fig.3g,h).Before Anc.3-11 duplicated,the protein ring a result,both Vma3 and Vmall are required in extant yeast to form a contained copies of only undifferentiated subunit Vma3/Vmall and complete ring with Vmal6. subunit 16.Immediately after Anc.3-11 duplicated,the two descend- To determine whether interaction interfaces were lost during evolu- ant subunits must have been functionally identical,so the protein ring tion,we engineered fusions of ancestral ring proteins to assess the could have assembled with many possible combinations of the two capacity of each to form the specific interfaces with other subunits descendants,including copies of only one of the descendant proteins. that are required for a functional complex.Because Anc.3-11 can This flexibility disappeared when Anc.3 lost the ancestral interface that complement the loss of both Vma3 and Vmall,we proposed that allowed it to interact with the anticlockwise side of Vmal6,and Anc.11 the Anc.3-11 subunit could participate in all three specific interaction lost the ability to interact with the clockwise side of Vmal6 and/or the interfaces,and that these capacities were then partitioned between anticlockwise side of Vma3.These complementary losses are sufficient Anc3 and Ancl1 after the duplication of Anc.3-11 (Fig.3a,b).To test to explain the specific arrangement of contemporary subunits in this hypothesis,we created six reciprocal gene fusions between yeast reconstructed and present-day fungal transmembrane rings. subunit Vmal6 and ancestral subunits Anc.3-11,Anc.3 and Anc.11 To establish the genetic basis for the partitioning of the functions of (Fig.3c and Supplementary Information,section 9).Each fusion con- Anc.3-11 between Vma3 and Vmall,we introduced historical muta- strains the structural position of subunits relative to subunit Vmal6, tions into Anc3.11 by directed mutagenesis and determined whether making it possible to determine which arrangements yield a functional they recapitulated the shifts in function that occurred during the evolu ring.As predicted,Anc.3-11 functioned on either side of Vmal6 tion of Anc.3 and Anc.11.The two phylogenetic branches leading from (Fig.3d),indicating that it could form all three interfaces P,Q and Anc.3-11 to Anc.3 and to Anc.11 contain 25 and 31 amino acid sub- R.By contrast,Anc.3 functioned when constrained to participate in stitutions,respectively,but only a subset of these are strongly con- interface Q with Vmal6 and interface P with Vma3;however,ring served in subunits Vma3 or Vmall from extant Fungi(Fig.4a).We function was lost when Anc.3 was constrained to form interface R with introduced each of these 'diagnostic'substitutions into Anc.3-11 and nc.161 映明 --11 llland I E 18W and V Duplication Vma16( Anc.X Anc.X Anc3-111 Anc 3-11 18■ Vma16 and ll and Plasmid Genotype YEPD CaCl 11 l and Iv Anc 16 I and V WT None - 34114164 Complementary losses Anc.3-11 duplication and Anc.3-11,[ma16-Anc.3-11] 34114164 complementary losses 18 Anc3-11,ne.3-11= ma16 34114164 Anc 3land IV Anc 11 ll and IV Vma16.Vma11 WT Anc.16 IV and V Vma3,Vma11- 164 Vma3.Vme11.[ma16Anc.31 164 34164 Anc.3.Vma11.IAnc.3Vma161 .--WT --114164 114164 114164 Vma3.[Anc. Figure 3Increasing complexity by complementary loss of interactions in form.For each experiment,expressed Vo subunits are listed.Tethered subunits the fungal Vo ring.a,Model of the ancestral three-paralogue ring,arranged as are in brackets and connected by a thick line.Cartoons show the constrained in extant yeast-.Unique intersubunit interfaces are labelled P,Q and R.Subunits location of the tethered subunit relative to Vmal6.Anc.3-11 can function on are colour-coded as in Fig.1.b,Model ofthe ancestral two-paralogue ring,before either side of Vmal6(d).Anc.3 can function only on the clockwise side of duplication of Anc.3-11.c,To constrain the location of specific subunits,gene Vmal6 (e).Anc.11 can function only on the anticlockwise side of Sc.16 fusions were constructed by tethering an ancestral subunit to either the amino- (f).g.Interfaces that are formed by Vo subunits before and after duplication and or carboxy-terminal side of yeast Vmal6.Roman numerals indicate the complementary loss of interfaces,based on the data in panels d-f.Red crosses locations of transmembrane helices (I,II,III,IV and V).d-f,Growth assays of indicate lost interfaces.h,Schematic of interfaces formed by Anc.3-11 that were yeast with fused Vo subunits identify the interfaces that ancestral subunits can lost in Anc.3 and Anc.11,based on data in panels d-f. I VOL 48119 JANUARY 2012 2012 Macmillan Publishers Limited.All rights reserved

extant yeast proteins have shown that the arrangement of subunits in the ring is constrained by the capacity of each subunit to form specific interfaces (which we labelled P, Q and R) with the other subunits24. Specifically, Vma11 is restricted to a single position between Vma16 and Vma3, because its clockwise interface can participate only in interface R with Vma16, and its anticlockswise interface can participate only in interface P with the clockwise side of Vma3 (Fig. 3). By contrast, copies of Vma3 occupy several positions in the ring, because they form interface P with other copies of Vma3 or Vma11, as well as interface Q with Vma16. However, Vma3 cannot form interface R with Vma16. As a result, both Vma3 and Vma11 are required in extant yeast to form a complete ring with Vma16. To determine whether interaction interfaces were lost during evolu￾tion, we engineered fusions of ancestral ring proteins to assess the capacity of each to form the specific interfaces with other subunits that are required for a functional complex. Because Anc.3-11 can complement the loss of both Vma3 and Vma11, we proposed that the Anc.3-11 subunit could participate in all three specific interaction interfaces, and that these capacities were then partitioned between Anc3 and Anc11 after the duplication of Anc.3-11 (Fig. 3a, b). To test this hypothesis, we created six reciprocal gene fusions between yeast subunit Vma16 and ancestral subunits Anc.3-11, Anc.3 and Anc.11 (Fig. 3c and Supplementary Information, section 9). Each fusion con￾strains the structural position of subunits relative to subunit Vma16, making it possible to determine which arrangements yield a functional ring. As predicted, Anc.3-11 functioned on either side of Vma16 (Fig. 3d), indicating that it could form all three interfaces P, Q and R. By contrast, Anc.3 functioned when constrained to participate in interface Q with Vma16 and interface P with Vma3; however, ring function was lost when Anc.3 was constrained to form interface R with Vma16 (Fig. 3e). Anc.11 functioned when constrained to participate in interface R with Vma16 and interface P with Vma3, but ring function was lost when Anc.11 was constrained to participate in interface Q with Vma16 and interface P with Vma3. This result indicates that Anc.11 lost the capacity to form one or both of these interfaces during its post-duplication divergence from Anc.3-11 (Fig. 3f). Taken together, these data indicate that the specificity of the ring arrangement and the obligate roles of Vma3 and Vma11 evolved by complementary loss of asymmetric interactions with other members of the ring (Fig. 3g, h). Before Anc.3-11 duplicated, the protein ring contained copies of only undifferentiated subunit Vma3/Vma11 and subunit 16. Immediately after Anc.3-11 duplicated, the two descend￾ant subunits must have been functionally identical, so the protein ring could have assembled with many possible combinations of the two descendants, including copies of only one of the descendant proteins. This flexibility disappeared when Anc.3 lost the ancestral interface that allowed it to interact with the anticlockwise side of Vma16, and Anc.11 lost the ability to interact with the clockwise side of Vma16 and/or the anticlockwise side of Vma3. These complementary losses are sufficient to explain the specific arrangement of contemporary subunits in reconstructed and present-day fungal transmembrane rings. To establish the genetic basis for the partitioning of the functions of Anc.3-11 between Vma3 and Vma11, we introduced historical muta￾tions into Anc3.11 by directed mutagenesis and determined whether they recapitulated the shifts in function that occurred during the evolu￾tion of Anc.3 and Anc.11. The two phylogenetic branches leading from Anc.3-11 to Anc.3 and to Anc.11 contain 25 and 31 amino acid sub￾stitutions, respectively, but only a subset of these are strongly con￾served in subunits Vma3 or Vma11 from extant Fungi (Fig. 4a). We introduced each of these ‘diagnostic’ substitutions into Anc.3-11 and P P P R P Q I II IV III I IV II III I II III IV I II III IV I II III IV I II III IV V b I II IV III I IV II III I II III IV I II III IV I II III IV I II III IV V P P P R P Q a Duplication Complementary losses g P,R P,Q I II III IV h P,R P,Q I II III IV P,R P,Q I II III IV II III IV V I IVIIIII N C I IIIII IV II IV VIII N C Cytosol Lumen Anc.X Vma16 R c f e YEPD CaCl2 d WT Plasmid Genotype Vma16, Vma11, Vma3 None Vma16, Vma11, Vma3 Vma3, Vma11 Anc.X Vma16 Q Anc.3-11, [Vma16 ▃Anc.3-11] Vma16, Vma11, Vma3 Vma3 Vma3, [Vma16 ▃ Anc.11] Vma3, [Anc.11 ▃ Vma16] Vma3, Vma11, [Vma16 ▃Anc.3] Anc.3, Vma11, [Anc.3 ▃ Vma16] Anc.3-11, [Anc.3-11▃ Vma16] WT WT Anc.3-11 duplication and complementary losses Anc.3 Anc.3 Anc.11 Anc.3 Anc.16 Anc.3 Anc.3 Anc.3-11 Anc.3-11 Anc.3-11 Anc.3-11 Anc.3-11 Anc.16 Anc.11 Anc.3-11 R --- P Q Anc.3-11 I and II Anc.16 II and III Anc.3-11 III and IV Anc.16 IV and V R P P Anc.3-11 I and II --- Q Q Anc.16 II and III R P P Anc.3-11 I and II Anc.3-11 III and IV Anc.16 IV and V Anc.3-11 III and IV Anc.3 I and II Anc.3 III and IV Anc.16 IV and V Anc.11 III and IV P P --- Q Anc.16 II and III R Anc.11 I and II 3Δ 11Δ 16Δ 3Δ 11Δ 16Δ 3Δ 11Δ 16Δ 16Δ 16Δ 3Δ 16Δ 11Δ 16Δ 11Δ 16Δ 11Δ 16Δ Figure 3 | Increasing complexity by complementary loss of interactions in the fungal V0 ring. a, Model of the ancestral three-paralogue ring, arranged as in extant yeast24. Unique intersubunit interfaces are labelled P, Q and R. Subunits are colour-coded as in Fig. 1. b,Model of the ancestral two-paralogue ring, before duplication of Anc.3-11. c, To constrain the location of specific subunits, gene fusions were constructed by tethering an ancestral subunit to either the amino￾or carboxy-terminal side of yeast Vma16. Roman numerals indicate the locations of transmembrane helices (I, II, III, IV and V)24. d–f, Growth assays of yeast with fused V0 subunits identify the interfaces that ancestral subunits can form. For each experiment, expressed V0 subunits are listed. Tethered subunits are in brackets and connected by a thick line. Cartoons show the constrained location of the tethered subunit relative to Vma16. Anc.3-11 can function on either side of Vma16 (d). Anc.3 can function only on the clockwise side of Vma16 (e). Anc.11 can function only on the anticlockwise side of Sc.16 (f). g, Interfaces that are formed by V0 subunits before and after duplication and complementary loss of interfaces, based on the data in panels d–f. Red crosses indicate lost interfaces. h, Schematic of interfaces formed by Anc.3-11 that were lost in Anc.3 and Anc.11, based on data in panels d–f. RESEARCH LETTER 362 | NATURE | VOL 481 | 19 JANUARY 2012 ©2012 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH b Plasmid Genotype YEPD CaCl, Plasmid None WT Anc3-11 None 34 Anc.11 None Anc.3-11 34 Anc.3 None Anc.3-11 V15F None Nane 34 None Anc.3-11 114 Anc.3-11 V15F 11A 155L A61S Y87S F109 YEPD An Plasmid Genotype T121y A122M None WT 132V V15A None 34 M22 Anc.3-11 34 825T M48 44 .Anc.3-11M22I 34 N88T H920 Anc.3-11 114 A120G 5 N159D Anc.3-11 M221 114 Figure 4 Genetic basis for functional differentiation of Anc.3 and Anc.11 table shows growth semiquantitatively from zero (none)to wild type a,Experimental analysis of historical amino acid replacements.The table lists (++++++)Bold mutations entirely or partly recapitulate the functional replacements that occurred on the branches leading from Anc.3-11 to Anc.11 evolution of Anc.11 and Anc.3.b,Replacement V15F abolishes the capacity of (yellow)or to Anc.3(blue)and that were subsequently conserved.Each derived Anc.3-11 to function as subunit 3 and enhances the capacity of Anc.3-11 to residue was introduced singly into Anc.3-11;the variant genes were function as subunit 11.c,Replacement M22I impairs the capacity of Anc.3-11 transformed into S.cerevisiae,and growth was assayed on elevated CaCl.The to function as subunit 11 without affecting its capacity to function as subunit 3. experimentally evaluated whether they recapitulated the loss by Anc.3 interaction screens and the phenotype of vmallA yeast provide no or Anc.11 of the capacity to complement Vma gene deletions.We evidence that Vmall evolved novel functions in addition to those that found that a single amino-acid replacement that occurred on the it inherited from Anc.3-11 in the Vo ring29 branch leading to Anc.11(V15F)abolished the capacity of Anc.3-11 We are aware of no other mechanistic analyses of a molecular to function as subunit 3;it also enhanced the ability of Anc.3-11 to machine's evolutionary trajectory,so the generality of our observations function as subunit 11 (Fig.4b).V15F is located in transmembrane is unknown.By definition,however,all molecular machines involve helix I,which participates in the P interface that our experiments differentiated parts in specific spatial orientations,and many such indicate may have been lost on the same branch (Fig.3 and complexes are entirely or partially composed of paralogous proteins2-. Supplementary Information,section 4).Conversely,a single historical In the evolution of any such assembly,additional paralogues could replacement(M221)on the branch leading to Anc.3 radically reduced become obligate components because of gene duplication and sub- the capacity of Anc.3-11 to function as subunit 11(Fig.4c).M22I is sequent mutations that cause specific interaction interfaces among also in transmembrane helix I,which participates in formation of the R them to degenerate. interface that was lost on this branch (Fig.3 and Supplementary This view of the evolution of molecular machines is related to recent Information,section 4).The Anc.3-11 M22I mutant retains some of models that explain other biological phenomena-such as the reten- the capacity of the ancestral protein to rescue growth in the Vmall- tion of large numbers of duplicate genes and mobile genetic elements deficient background,suggesting that other mutations also contributed within genomes-as the product of degenerative processes acting on to the functional evolution of Vma3.One other historical mutation modular biological systems27.Although mutations that enhanced the (N88T)on this branch also impaired the capacity of Anc.3-11 to func- functions of individual ring components may have occurred during tion as subunit 11,but it reduced the capacity of the protein to function evolution,our data indicate that simple degenerative mutations are as Vma3 as well,suggesting that epistatic interactions with other residues sufficient to explain the historical increase in complexity of a crucial allow this mutation to be tolerated in Anc.3 and its descendants.Several molecular machine.There is no need to invoke the acquisition of of the replacements on the branch leading to Anc.11 show a similar novel'functions caused by low-probability mutational combinations pattern,reducing the capacity of the protein to replace Vma3,indi- cating that these historical replacements function better together than METHODS SUMMARY in isolation. Ancestral protein sequences were inferred using maximum-likelihood phylogenetics How complexity and novel functions evolve has been a longstand- from an alignment of 139 protein sequences of extant subunits 3,11 and 16 from ing question in evolutionary biology,because mutations that com- Amoebozoa,Apicomplexa,Metazoa,Choanoflagellida and Fungi.Ancestral genes promise existing functions are far more frequent than those that were synthesized,cloned into yeast expression vectors and tested for complementa- generate new ones2s.Our results indicate that the architectural com- tion in various S.cerevisiae mutants.V-ATPase function was assayed by growth tests plexity of molecular assemblies can evolve because of a few simple, on medium buffered with CaCl,,as described previously.Steady-state levels of relatively high-probability mutations that degrade ancestral interfaces Vphl were determined by western blot.Quinacrine staining and Vph1-GFP (green but leave other functions intact.The specific roles of subunits Vma3 fluorescent protein)fusion constructs were visualized by fluorescence microscopy. and Vmall seem to have been acquired when duplicated genes lost Full Methods and any associated references are available in the online version of some,but not all,of the capacity of the ancestral protein to participate the paper at www.nature.com/nature. in interactions with copies of itself and another protein required for proper ring assembly.Because complementary losses occurred in both Received 21 September;accepted 21 November 2011 lineages,the two descendant subunits became obligate components, Published online 9 January 2012. and the complexity ofthe ring increased.It is possible that specialization 1.Forgac,M.Vacuolar ATPases:rotary proton pumps in physiology and of the duplicated subunits allowed increases in fitness,but genome-wide pathophysiology.Nature Rev.Mol.Cell Biol.8,917-929 (2007). 19 JANUARY 2012 VOL 481 NATURE 363 2012 Macmillan Publishers Limited.All rights reserved

experimentally evaluated whether they recapitulated the loss by Anc.3 or Anc.11 of the capacity to complement Vma gene deletions. We found that a single amino-acid replacement that occurred on the branch leading to Anc.11 (V15F) abolished the capacity of Anc.3-11 to function as subunit 3; it also enhanced the ability of Anc.3-11 to function as subunit 11 (Fig. 4b). V15F is located in transmembrane helix I, which participates in the P interface that our experiments indicate may have been lost on the same branch (Fig. 3 and Supplementary Information, section 4). Conversely, a single historical replacement (M22I) on the branch leading to Anc.3 radically reduced the capacity of Anc.3-11 to function as subunit 11 (Fig. 4c). M22I is also in transmembrane helix I, which participates in formation of the R interface that was lost on this branch (Fig. 3 and Supplementary Information, section 4). The Anc.3-11 M22I mutant retains some of the capacity of the ancestral protein to rescue growth in the Vma11- deficient background, suggesting that other mutations also contributed to the functional evolution of Vma3. One other historical mutation (N88T) on this branch also impaired the capacity of Anc.3-11 to func￾tion as subunit 11, but it reduced the capacity of the protein to function asVma3 as well, suggesting that epistatic interactions with other residues allow this mutation to be tolerated in Anc.3 and its descendants. Several of the replacements on the branch leading to Anc.11 show a similar pattern, reducing the capacity of the protein to replace Vma3, indi￾cating that these historical replacements function better together than in isolation. How complexity and novel functions evolve has been a longstand￾ing question in evolutionary biology25–27, because mutations that com￾promise existing functions are far more frequent than those that generate new ones28. Our results indicate that the architectural com￾plexity of molecular assemblies can evolve because of a few simple, relatively high-probability mutations that degrade ancestral interfaces but leave other functions intact. The specific roles of subunits Vma3 and Vma11 seem to have been acquired when duplicated genes lost some, but not all, of the capacity of the ancestral protein to participate in interactions with copies of itself and another protein required for proper ring assembly. Because complementary losses occurred in both lineages, the two descendant subunits became obligate components, and the complexity of the ring increased. It is possible that specialization of the duplicated subunits allowed increases infitness, but genome-wide interaction screens and the phenotype of vma11D yeast provide no evidence that Vma11 evolved novel functions in addition to those that it inherited from Anc.3-11 in the V0 ring29. We are aware of no other mechanistic analyses of a molecular machine’s evolutionary trajectory, so the generality of our observations is unknown. By definition, however, all molecular machines involve differentiated parts in specific spatial orientations, and many such complexes are entirely or partially composed of paralogous proteins2–8. In the evolution of any such assembly, additional paralogues could become obligate components because of gene duplication30 and sub￾sequent mutations that cause specific interaction interfaces among them to degenerate. This view of the evolution of molecular machines is related to recent models that explain other biological phenomena—such as the reten￾tion of large numbers of duplicate genes and mobile genetic elements within genomes—as the product of degenerative processes acting on modular biological systems27. Although mutations that enhanced the functions of individual ring components may have occurred during evolution, our data indicate that simple degenerative mutations are sufficient to explain the historical increase in complexity of a crucial molecular machine. There is no need to invoke the acquisition of ‘novel’ functions caused by low-probability mutational combinations. METHODS SUMMARY Ancestral protein sequences were inferred using maximum-likelihood phylogenetics from an alignment of 139 protein sequences of extant subunits 3, 11 and 16 from Amoebozoa, Apicomplexa, Metazoa, Choanoflagellida and Fungi. Ancestral genes were synthesized, cloned into yeast expression vectors and tested for complementa￾tion in various S. cerevisiae mutants. V-ATPase function was assayed by growth tests on medium buffered with CaCl2, as described previously31. Steady-state levels of Vph1 were determined by western blot. Quinacrine staining and Vph1–GFP (green fluorescent protein) fusion constructs were visualized by fluorescence microscopy. Full Methods and any associated references are available in the online version of the paper at www.nature.com/nature. Received 21 September; accepted 21 November 2011. Published online 9 January 2012. 1. Forgac, M. Vacuolar ATPases: rotary proton pumps in physiology and pathophysiology. Nature Rev. Mol. Cell Biol. 8, 917–929 (2007). a b c CaCl Plasmid Genotype YEPD 2 Plasmid Genotype YEPD WT Anc.3-11 Anc.3-11 V15F Anc.3-11 Anc.3-11 V15F Anc.3-11 Anc.3-11 M22I Anc.3-11 Anc.3-11 M22I None None None None Plasmid Yeast growth in 3 Yeast growth Anc.3-11 ++++ ++ Anc.11 None ++++++ Anc.3 ++++++ None None None None V15F None ++++ M16A ++ ++ V38I ++++ ++ A42G +++ ++ V45T ++ ++ M46F ++ ++ I55L +++++ +++ A61S ++++ ++ Y87S ++ ++ F108Y + ++ T121Y ++++ ++ A122M + ++ I132V +++++ ++ V15A + +++ M22I +++++ + S25T +++ ++ M46L +++ ++ N88T ++ + H92Q ++ ++ A120G ++ ++ N159D ++ ++ Anc.3-11 Anc.3 Anc.11 Δ in 11Δ 3Δ 3Δ 3Δ 11Δ 11Δ WT 3Δ 3Δ 3Δ 11Δ 11Δ CaCl2 Figure 4 | Genetic basis for functional differentiation of Anc.3 and Anc.11. a, Experimental analysis of historical amino acid replacements. The table lists replacements that occurred on the branches leading from Anc.3-11 to Anc.11 (yellow) or to Anc.3 (blue) and that were subsequently conserved. Each derived residue was introduced singly into Anc.3-11; the variant genes were transformed into S. cerevisiae, and growth was assayed on elevated CaCl2. The table shows growth semiquantitatively from zero (none) to wild type (111111). Bold mutations entirely or partly recapitulate the functional evolution of Anc.11 and Anc.3. b, Replacement V15F abolishes the capacity of Anc.3-11 to function as subunit 3 and enhances the capacity of Anc.3-11 to function as subunit 11. c, Replacement M22I impairs the capacity of Anc.3-11 to function as subunit 11 without affecting its capacity to function as subunit 3. LETTER RESEARCH 19 JANUARY 2012 | VOL 481 | NATURE | 363 ©2012 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER 2 Pallen,M.J.Matzke,N.J.From the origin of species to the origin of bacterial 20.Taylor,J.W.Berbee,M.L.Dating divergences in the fungal tree of life:review and flagella.Nature Rev.Microbiol.4,784-790 (2006). new analyses.Mycologia98,838-849(2006). 3.Liu,R.Ochman.H.Stepwise formation of the bacterial flagellar system.Proc.Natl 21.Yang.Z.Kumar,S.Nei,M.Anew method of inference of ancestral nucleotide and Acad.Sci.USA104,7116-7121(200. amino acid sequences.Genetics 141,1641-1650(1995) 4.Mulkidianian.A.Y..Makarova.K.S..Galperin,M.Y.Koonin,E V.Inventing the 22.Kane,P.M.The where,when,and how of organelle acidification by the yeast dynamo machine:the evolution of the F-type and V-type ATPases.Nature Rev. vacuolar H+-ATPase.Microbiol.Mol.Biol.Rev.70,177-191 (2006). Microbiol5,892-899(2007). 23.Hirata,R.,Graham,LA,Takatsuki,A,Stevens,T.H.Anraku,Y.Vmal1 and 5. Dolezal,P.Likic,V.Tachezy,J.&Lithgow,T.Evolution of the molecular machines vma16 encode second and third proteolipid subunits of the Saccharomyces for protein import into mitochondria.Science 313,314-318(2006). cerevisiae vacuolar membrane H+-ATPase.J.Biol.Chem.272,4795-4803 6. Clements,Aetal.The reducible complexity of a mitochondrial molecular machine. (1997. Proc.Natl Acad.Sci.USA 106,15791-15795(2009). 24. Wang.Y.Cipriano,D.J.&Forgac,M.Arrangement of subunits in the proteolipid 7. Archibald,J.M.Logsdon,J.M.Jr Doolittle,W.F.Origin and evolution of ring of the V-ATPase.J.Biol.Chem.282,34058-34065(2007). eukaryotic chaperonins:phylogenetic evidence for ancient duplications in CCT Ohno.S.Evolution by Gene Duplication(Springer,1970) genes.MoL Biol.Evol 17,1456-1466(2000). 26.Jacob,F.Evolution and tinkering.Science 196,1161-1166(1977) 8. Gabaldon,T..Rainey.D.Huynen,M.A.Tracing the evolution of a large protein 27.Lynch,M.The frailty of adaptive hypotheses for the origins of organismal complex in the eukaryotes,NADH:ubiquinone oxidoreductase(complex I).J.Mol. complexity.Proc.Natl Acad.Sci.USA 104,8597-8604(2007) Bi0l348.857-870(2005). 28.Hietpas,R.T.Jensen,J.D.&Bolon,D.N.Experimental illumination of a fitness 9 Thornton,J.W.Resurrecting ancient experimental analysis of extinct landscape.Proc.Nat/Acad.Sci.USA 108,7896-7901 (2011). molecules.Nature Rev.Genet.5,366-375 (2004). 29.Tong.A.H.Y.etal Global mapping ofthe yeast genetic interaction network.Science 10.Liberles,D.(ed.)Ancestral Seguence Reconstruction(Oxford Univ.Press,2007) 303,808-813(2004) 11.Harms,M.J.Thornton,J.W.Analyzing protein structure and function using 30.Pereira-Leal,J.B.Levy,E.D.,Kamp,C.Teichmann,S.A Evolution of protein ancestral gene reconstruction.Curr.Opin.Struct Biol.20,360-366(2010). complexes by duplication of homomeric interactions.Genome Biol.8,R51(2007) 12.Frattini.A.et al.Defects in TCIRGI subunit of the vacuolar proton pump are 31.Ryan,M.,Graham,LA.Stevens,T.H.Voalp functions in V-ATPase assembly in responsible r a subset of human autosomal recessive osteopetrosis.Nature the yeast endoplasmic reticulum.Mol Biol.Cell 19,5131-5142 (2008). Genet25,343-346(2000). Supplementary Information is linked to the online version of the paper at 13.Perez-Sayans,M.Somoza-Martin,J.M.,Barros-Angueira,F.Rey,J.M.Garcia www.nature.com/nature. Garcia,A.V-ATPase inhibitors and implication in cancer treatment.Cancer Treat Rev.35,707-713(2009) Acknowledgements This study was supported by National Institutes of Health(NIH) 14.Xu.L et al Inhibition of host vacuolar H-ATPase activity by a Legionella grants RO1-GM081592(to J.W.T.)and RO1-GM38006(to T.H.S.),National Science pneumophila effector.PLoS Pathog.6,e1000822(2010). Foundation (NSF)grants IOB-0546906 (to J.W.T.)and DEB-0516530 (to J.W.T.).NIH 15.Hirata,T.et al.Subunit rotation of vacuolar-type proton pumping ATPase:relative Genetics Training grant T32-GM007257(to G.C.F.),NSF IGERT grant DGE-9972830 rotation of the g and c subunits.J.Biol Chem.278,23714-23719(2003). (to V.H.-S.)and the Howard Hughes Medical Institute (J.W.T.).We thank L Graham, 16.Imamura,H.et al Rotation scheme of Vi-motor is different from that of Fi-motor. G.Butler and B.Houser for generating yeast strains and other assistance.We thank Proc.Natl Acad.Sci USA 102,17929-17933(2005). members of the Stevens and Thomnton laboratories for helpful comments 17.Powell,B.Graham,L.A.Stevens,T.H.Molecular characterization of the yeast vacuolar H+-ATPase proton pore.J.Biol.Chem.275,23654-23660 (2000). Author Contributions V.H.-S.performed the phylogenetic analysis and statistical reconstructions.G.C.F.performed functional experiments.All authors conceived the 18.Umemoto,N.,Yoshihisa,T.Hirata,R.Anraku,Y.Roles of the VMA3 gene product, experiments,interpreted the results and wrote the paper. subunit c of the vacuolar membrane H-ATPase on vacuolar acidification and protein transport.A study with VMA3-disrupted mutants of Saccharomyces Author Information Reprints and permissions information is available at cerevisiae.1 BioL Chem.265,18447-18453(1990). www.nature.com/reprints.The authors declare no competing financial interests. 19.Umemoto,N.,Ohya,Y.Anraku,Y.VMA11,a novel gene that encodes a putative Readers are welcome to comment on the online version of this article at proteolipid,is indispensable for expression of yeast vacuolar membrane www.nature.com/nature.Correspondence and requests for materials should be H*-ATPase activity.J.Biol Chem.266,24526-24532(1991). addressed to J.W.T.(joet@uoregon.edu). 364 NATURE VOL 481|19 JANUARY 2012 2012 Macmillan Publishers Limited.All rights reserved

2. Pallen, M. J. & Matzke, N. J. From the origin of species to the origin of bacterial flagella. Nature Rev. Microbiol. 4, 784–790 (2006). 3. Liu, R. & Ochman, H. Stepwise formation of the bacterial flagellar system. Proc. Natl Acad. Sci. USA 104, 7116–7121 (2007). 4. Mulkidjanian, A. Y., Makarova, K. S., Galperin, M. Y. & Koonin, E. V. Inventing the dynamo machine: the evolution of the F-type and V-type ATPases. Nature Rev. Microbiol. 5, 892–899 (2007). 5. Dolezal, P., Likic, V., Tachezy, J. & Lithgow, T. Evolution of the molecular machines for protein import into mitochondria. Science 313, 314–318 (2006). 6. Clements, A. et al.The reducible complexity of amitochondrialmolecularmachine. Proc. Natl Acad. Sci. USA 106, 15791–15795 (2009). 7. Archibald, J. M., Logsdon, J. M. Jr & Doolittle, W. F. Origin and evolution of eukaryotic chaperonins: phylogenetic evidence for ancient duplications in CCT genes. Mol. Biol. Evol. 17, 1456–1466 (2000). 8. Gabaldo´n, T., Rainey, D. & Huynen, M. A. Tracing the evolution of a large protein complex in the eukaryotes, NADH:ubiquinone oxidoreductase (complex I). J. Mol. Biol. 348, 857–870 (2005). 9. Thornton, J. W. Resurrecting ancient genes: experimental analysis of extinct molecules. Nature Rev. Genet. 5, 366–375 (2004). 10. Liberles, D. (ed.) Ancestral Sequence Reconstruction (Oxford Univ. Press, 2007). 11. Harms, M. J. & Thornton, J. W. Analyzing protein structure and function using ancestral gene reconstruction. Curr. Opin. Struct. Biol. 20, 360–366 (2010). 12. Frattini, A. et al. Defects in TCIRG1 subunit of the vacuolar proton pump are responsible for a subset of human autosomal recessive osteopetrosis. Nature Genet. 25, 343–346 (2000). 13. Pe´rez-Saya´ns, M., Somoza-Martı`n, J. M., Barros-Angueira, F., Rey, J. M. & Garcı`a￾Garcı`a, A. V-ATPase inhibitors and implication in cancer treatment. Cancer Treat. Rev. 35, 707–713 (2009). 14. Xu, L. et al. Inhibition of host vacuolar H1-ATPase activity by a Legionella pneumophila effector. PLoS Pathog. 6, e1000822 (2010). 15. Hirata, T. et al. Subunit rotation of vacuolar-type proton pumping ATPase: relative rotation of the g and c subunits. J. Biol. Chem. 278, 23714–23719 (2003). 16. Imamura, H. et al. Rotation scheme of V1-motor is different from that of F1-motor. Proc. Natl Acad. Sci. USA 102, 17929–17933 (2005). 17. Powell, B., Graham, L. A. & Stevens, T. H. Molecular characterization of the yeast vacuolar H1-ATPase proton pore. J. Biol. Chem. 275, 23654–23660 (2000). 18. Umemoto, N., Yoshihisa, T., Hirata, R. & Anraku, Y. Roles of the VMA3 gene product, subunit c of the vacuolar membrane H1-ATPase on vacuolar acidification and protein transport. A study with VMA3-disrupted mutants of Saccharomyces cerevisiae. J. Biol. Chem. 265, 18447–18453 (1990). 19. Umemoto, N., Ohya, Y. & Anraku, Y. VMA11, a novel gene that encodes a putative proteolipid, is indispensable for expression of yeast vacuolar membrane H1-ATPase activity. J. Biol. Chem. 266, 24526–24532 (1991). 20. Taylor, J. W. & Berbee, M. L. Dating divergences in the fungal tree of life: review and new analyses. Mycologia 98, 838–849 (2006). 21. Yang, Z., Kumar, S. & Nei, M. A new method of inference of ancestral nucleotide and amino acid sequences. Genetics 141, 1641–1650 (1995). 22. Kane, P. M. The where, when, and how of organelle acidification by the yeast vacuolar H1-ATPase. Microbiol. Mol. Biol. Rev. 70, 177–191 (2006). 23. Hirata, R., Graham, L. A., Takatsuki, A., Stevens, T. H. & Anraku, Y. Vma11 and vma16 encode second and third proteolipid subunits of the Saccharomyces cerevisiae vacuolar membrane H1-ATPase. J. Biol. Chem. 272, 4795–4803 (1997). 24. Wang, Y., Cipriano, D. J. & Forgac, M. Arrangement of subunits in the proteolipid ring of the V-ATPase. J. Biol. Chem. 282, 34058–34065 (2007). 25. Ohno, S. Evolution by Gene Duplication (Springer, 1970). 26. Jacob, F. Evolution and tinkering. Science 196, 1161–1166 (1977). 27. Lynch, M. The frailty of adaptive hypotheses for the origins of organismal complexity. Proc. Natl Acad. Sci. USA 104, 8597–8604 (2007). 28. Hietpas, R. T., Jensen, J. D. & Bolon, D. N. Experimental illumination of a fitness landscape. Proc. Natl Acad. Sci. USA 108, 7896–7901 (2011). 29. Tong, A. H. Y. et al. Globalmapping of the yeast genetic interaction network. Science 303, 808–813 (2004). 30. Pereira-Leal, J. B., Levy, E. D., Kamp, C. & Teichmann, S. A. Evolution of protein complexes by duplication of homomeric interactions. Genome Biol. 8,R51 (2007). 31. Ryan, M., Graham, L. A. & Stevens, T. H. Voa1p functions in V-ATPase assembly in the yeast endoplasmic reticulum. Mol. Biol. Cell 19, 5131–5142 (2008). Supplementary Information is linked to the online version of the paper at www.nature.com/nature. Acknowledgements This study was supported by National Institutes of Health (NIH) grants R01-GM081592 (to J.W.T.) and R01-GM38006 (to T.H.S.), National Science Foundation (NSF) grants IOB-0546906 (to J.W.T.) and DEB-0516530 (to J.W.T.), NIH Genetics Training grant T32-GM007257 (to G.C.F.), NSF IGERT grant DGE-9972830 (to V.H.-S.) and the Howard Hughes Medical Institute (J.W.T.). We thank L. Graham, G. Butler and B. Houser for generating yeast strains and other assistance. We thank members of the Stevens and Thornton laboratories for helpful comments. Author Contributions V.H.-S. performed the phylogenetic analysis and statistical reconstructions. G.C.F. performed functional experiments. All authors conceived the experiments, interpreted the results and wrote the paper. Author Information Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this article at www.nature.com/nature. Correspondence and requests for materials should be addressed to J.W.T. (joet@uoregon.edu). RESEARCH LETTER 364 | NATURE | VOL 481 | 19 JANUARY 2012 ©2012 Macmillan Publishers Limited. All rights reserved

ETTER RESEARCH METHODS cloned into pCR4Blunt-TOPO (Invitrogen).When necessary,a modified In silico reconstruction of ancestral protein sequences.Vo complex subunits Quikchange protocol4?was used to introduce point mutations before the gene Vma3,Vmall and Vmal6 are sometimes referred to as subunits c,c'and c"in the was subcloned into a yeast vector(pRS316 or pRS415).To generate pGF502, literature.We searched GenBank for all eukaryote V-ATPase Vo ring sequences sequence from codon 31 to the stop codon of Anc.16 was amplified with the (Supplementary Information,section 1).Our query returned subunit 3,11 and 16 ADH:Nat'cassette from pGF139,cloned into TOPO,and in vivo ligated down- protein sequences for 26 species in Fungi,and subunit 3 and 11 sequences for 35 stream of the VMA16 promoter (including a start codon)in pRS415. species in Metazoa,Amoebozoa and Apicomplexa.We aligned the sequences A triple-fragment in vivo ligation was used to generate pGF646-pGF651. using PRANK v0.081202(refs 32,33).We selected the best-fit model (WAG with Gapped vector containing the VMA16 promoter was transformed into yeast with gamma-distributed rate variation and a proportion of invariant sites)using the two PCR fragments of the ring genes to be fused.For pGF646,the coding region of Akaike Information Criterion as implemented in PROTTEST With this (1)VMA16(without codons 2-41)and (2)the coding region of Anc.11 (without model,we used PhyML v3.0 to infer the maximum likelihood topology,branch codons 2-5 were amplified by PCR.The proteolipid on the C-terminal portion of lengths and model parameters".We optimized the topology using the best result the gene fusion also contained the ADH terminator and Nar'cassette;the amp- from nearest-neighbour interchange and subtree pruning and regrafting;we lified products contained PCR tails with homology to link the genes to both the optimized all other free parameters using the default hill-climbing algorithm in gapped vector and to each other.Gene fusions were modelled after the experi- PhyML.Phylogenetic support was calculated as the approximate likelihood ratio mental design of Wang et al.(2007)in which the lumenal protein sequence (converted from the approximate likelihood ratio statistic (aLRS)for branches linking the two proteolipids was designed to be exactly 14 amino acids.To meet reported by PhyML,using the equation aLR exp[aLRS/2])and as the likelihood these criteria,additional amino acids were inserted into the following vectors ratio-based SH-like branch supports".Nematoda subunit 3 and 11 sequences linking the two subunits:pGF646 (Thr-Arg-Val-Asp),pGF648,pGF650 (Thr- were connected by a very long branch basal to the Chromalveolata lineages. Arg),pGF649,pGF651(Gly-Ser). This result is inconsistent with the expectation that Nematoda are animals,so Yeast strains that were used are listed in Supplementary Information,section 2. we excluded Nematoda data from further downstream analysis. Strains containing deletion cassettes other than KanR4s were constructed by PCR We inferred ML ancestral states and posterior probability distributions at each amplifying the Hyg"or Nar'cassette from pAG32 or pAG25,respectively,with site for all ancestral nodes in the ML phylogeny using our own set of Python scripts, primer tails with homology to flanking sequences to the VMA1l or VMA16 loci called Lazarus,which wraps PAML version 4.1(ref.39).Lazarus parsimoniously 114::Kan"and 164::Kan"strains(SF838-1Da)were transformed with the Hyg' places ancestral gap characters according to Fitch's algorithm.We characterized and Nar PCR fragments,respectively,and selected for drug resistance.The the overall support for Anc.3-11,Anc.16,Anc.3 and Anc.11 by binning the 114:Hyg"locus was amplified and transformed into LGY113 (to create posterior probability of the ML state at each site into 5%-sized bins and then LGY125)and LGY115 (to create LGY124).This was repeated with the counting the proportion of total sites within each bin (Supplementary Informa- 164::Nat locus to create LGY139 and LGY143. tion,section 2). Yeast Growth Assays.Yeast were grown in liquid culture,diluted fivefold and Robustness to alignment uncertainty.To assess the robustness of ancestral spotted onto YEPD media buffered to pH 5.0 or yeast extract peptone dextrose reconstructions to alignment uncertainty,we performed alignment using four media containing 25 mM (Figs 2,3,4)or 30 mM CaCl (Fig.2f). algorithms:CLUSTAL version 2.0.10 (ref.41),MUSCLE v3.7(ref.42),AMAP Whole-cell extract preparation and immunoblotting.Yeast extracts and west- v2.2 (ref.43),and PRANK v0.081202 (refs 32,33).We then inferred the ML ern blots were performed as previously described.Antibodies that were used in phylogeny and branch lengths for each alignment,using the methods described this study included monoclonal primary anti-HA (Sigma-Aldrich),anti-Dpml above.The resultant alignments varied in length from 347 sites (using CLUSTAL) (5C5;Invitrogen)and secondary horseradish-conjugated anti-mouse antibody to 683 sites(using PRANK),but all four alignments yielded the same ML topology (Jackson ImmunoResearch Laboratory,West Grove,Pennsylvania,USA). with nearly identical ML branch lengths. Fluorescence microscopy.Staining with quinacrine was performed as previously To determine which alignment algorithm yields the most accurate ancestral infer- described.The cell wall (shown in red)was visualized using concanavalin A ences under V-ATPase phylogenetic conditions,we simulated sequences across the tetramethylrhodamine (Invitrogen).Microscopy images were obtained using an V-ATPase MLphylogeny using insertion and deletion rates ranging from 0.0 to0.1 Axioplan 2 fluorescence microscope (Carl Zeiss).A X100 objective,AxioVision indels per site.For each indel rate,we generated ten random unique indel-free software (Carl Zeiss)and Adobe Photoshop Creative Suite (v.8.0)were used. ancestral sequences of 400 amino acids in length and then used INdelible#to 32.Loytynoja,A.Goldman,N.An algorithm for progressive multiple alignment of simulate the ancestral sequence evolving along the branches of our ML phylogeny sequences with insertions.Proc.NatlAcad Sci.USA 102,10557-10562(2005). under the conditions of the WAG model with a proportion of invariant sites (+I) 33. Loytynoja,A.Goldman,N.Phylogeny-aware gap placement prevents errors in and a discrete gamma distribution of evolutionary rates (+G)with indel events sequence alignment and evolutionary analysis.Science 320,1632-1635(2008) randomly injected according to the specified indel rate.The size of each indel event 34. Whelan,S.Goldman,N.A general empirical model of protein evolution derived was drawn from a Zipfian distribution with coefficient equal to 1.1 and the maximum from multiple protein families using a maximum-likelihood approach.Mol Biol. Evol18.691-699(2001). length limited to 10 amino acids.We aligned the descendant sequences of each 35.Abascal,F.Zardoya,R.&Posada,D.Prottest:selection of best-fit models of protein replicate using AMAP,CLUSTAL,MUSCLE and PRANK.For each alignment, evolution.Bioinformatics 21,2104-2105(2005) we inferred the ML topology,branch lengths and model parameters using the 36.Guindon,S.Gascuel,O.A simple,fast,and accurate algorithm to estimate large methods described above.We used Lazarus to reconstruct all of the ancestral states, phylogenies by maximum likelihood.Syst Biol.52,696-704(2003). and queried Lazarus for the most-recent shared ancestor for opisthokont subunit 37.Anisimova,M.Gascuel,O.Approximate likelihood-ratio test for branches:A fast, 3/11 and opisthokont subunit 16 sequences.We measured the error of ancestral accurate,and powerful alternative.Syst Biol 55,539-552 (2006). 38.Aguinaldo,A.M.Aetal.Evidence for a clade of nematodes,arthropods,and other reconstructions as the proportion of ancestral sites that incorrectly contained an moulting animals.Nature 387,489-493(1997). indel character(see Supplementary Information,section 6). 39. Yang,Z.PAML 4:Phylogenetic analysis by maximum likelihood.Mol.Biol.Evol.24, Plasmids and yeast strains.Bacterial and yeast manipulations were performed 1586-1591(2007) using standard laboratory protocols for molecular biology.Plasmids that were 40.Fitch,W.M.Toward defining the course of evolution:minimum change for a used are listed in Supplementary Information,section 5.Ancestral sequences specific tree topology.Syst Zool 20,406-416(1971). 41.Thompson,J.D.,Higgins,D.G.Gibson,T.J.CLUSTALW:improving the sensitivity (pGF140,pGF139,pGF506 and pGF508)were synthesized by GenScript with a of progressive multiple sequence alignment through sequence weighting yeast codon bias.Triple haemagglutinin epitope tags were included before each position specific gap penalties and weight matrix choice.Nucleic Acids Res.22 stop codon.The Anc.3-11,Anc.16,Anc.3 and Anc.11 genes were subcloned to 4673-4680(1994) single-copy,CEN-based yeast vectors.The ADH terminator sequence (247 base 42.Edgar,R.C.MUSCLE:multiple sequence alignment with high accuracy and high pairs (bp))and Nat'drug resistance marker were amplified using polymerase. throughput.Nucleic Acids Res.32,1792-1797(2004). chain-reaction (PCR)containing 40-bp tails homologous to the 3'end of each 43.Do,C.B.,Mahabhashyam,M.S.Brudno,M.Batzoglou,S.ProbCons: Probabilistic consistency-based multiple sequence alignment.Genome Res.15, coding region and vector sequence.Vectors were gapped,co-transformed into 330-340(2005). SF838-1Do yeast with PCR fragments and cells were selected for Nat.A second 44.Fletcher,W. &Yang.Z.Indelible:a flexible simulator of biological sequence round of in vivo ligation was used to place the ancestral genes under 500 bp of the evolution.MoL Biol.Evol.26,1879-1888 (2009). VMA3 or VMA16 promoters to create pGF140 and pGF139,respectively.For 45.Sambrook,J.Russel,D.W.Molecular Cloning:A Laboratory Manual 3rd edn(Cold vectors pGF240,pGF241,1pGF252,pGF253,pGF503-pGF508,pGF510, Spring Harbor Laboratory Press,2001). 46.Goldstein,A.L.McCuster,J.H.Three new dominantdrug resistance cassettes for pGF512-pGF515,pGF517-pGF519,pGF521,pGF523,pGF528,pGF529, gene disruption in Saccharomyces cerevisiae.Yeast 15,1541-1553(1999). pGF531,pGF534-pGF537 and pGF542,the relevant locus(Anc.3-11,Anc.16 or 47.Zheng.LBaumann,U.&Reymond,.L.An efficient one-step site-directed and Anc.3)was PCR amplified with 5'and 3'untranslated flanking sequence and site-saturation mutagenesis protocol.Nucleic Acids Res.32,e115(2004) 2012 Macmillan Publishers Limited.All rights reserved

METHODS In silico reconstruction of ancestral protein sequences. V0 complex subunits Vma3, Vma11 and Vma16 are sometimes referred to as subunits c, c9 and c0 in the literature. We searched GenBank for all eukaryote V-ATPase V0 ring sequences (Supplementary Information, section 1). Our query returned subunit 3, 11 and 16 protein sequences for 26 species in Fungi, and subunit 3 and 11 sequences for 35 species in Metazoa, Amoebozoa and Apicomplexa. We aligned the sequences using PRANK v0.081202 (refs 32, 33). We selected the best-fit model (WAG with gamma-distributed rate variation and a proportion of invariant sites) using the Akaike Information Criterion as implemented in PROTTEST34,35. With this model, we used PhyML v3.0 to infer the maximum likelihood topology, branch lengths and model parameters36. We optimized the topology using the best result from nearest-neighbour interchange and subtree pruning and regrafting; we optimized all other free parameters using the default hill-climbing algorithm in PhyML. Phylogenetic support was calculated as the approximate likelihood ratio (converted from the approximate likelihood ratio statistic (aLRS) for branches reported by PhyML, using the equation aLR 5 exp[aLRS/2]) and as the likelihood ratio-based SH-like branch supports37. Nematoda subunit 3 and 11 sequences were connected by a very long branch basal to the Chromalveolata lineages. This result is inconsistent with the expectation that Nematoda are animals38, so we excluded Nematoda data from further downstream analysis. We inferred ML ancestral states and posterior probability distributions at each site for all ancestral nodes in the ML phylogeny using our own set of Python scripts, called Lazarus, which wraps PAML version 4.1 (ref. 39). Lazarus parsimoniously places ancestral gap characters according to Fitch’s algorithm40. We characterized the overall support for Anc.3-11, Anc.16, Anc.3 and Anc.11 by binning the posterior probability of the ML state at each site into 5%-sized bins and then counting the proportion of total sites within each bin (Supplementary Informa￾tion, section 2). Robustness to alignment uncertainty. To assess the robustness of ancestral reconstructions to alignment uncertainty, we performed alignment using four algorithms: CLUSTAL version 2.0.10 (ref. 41), MUSCLE v3.7 (ref. 42), AMAP v2.2 (ref. 43), and PRANK v0.081202 (refs 32, 33). We then inferred the ML phylogeny and branch lengths for each alignment, using the methods described above. The resultant alignments varied in length from 347 sites (using CLUSTAL) to 683 sites (using PRANK), but all four alignments yielded the same ML topology with nearly identical ML branch lengths. To determine which alignment algorithm yields the most accurate ancestral infer￾ences under V-ATPase phylogenetic conditions, we simulated sequences across the V-ATPase ML phylogeny using insertion and deletion rates ranging from 0.0 to 0.1 indels per site. For each indel rate, we generated ten random unique indel-free ancestral sequences of 400 amino acids in length and then used INdelible44 to simulate the ancestral sequence evolving along the branches of our ML phylogeny under the conditions of the WAG model with a proportion of invariant sites (1I) and a discrete gamma distribution of evolutionary rates (1G) with indel events randomly injected according to the specified indel rate. The size of each indel event was drawnfrom a Zipfian distributionwith coefficient equal to 1.1 and themaximum length limited to 10 amino acids. We aligned the descendant sequences of each replicate using AMAP, CLUSTAL, MUSCLE and PRANK. For each alignment, we inferred the ML topology, branch lengths and model parameters using the methods described above. We used Lazarus to reconstruct all of the ancestral states, and queried Lazarus for the most-recent shared ancestor for opisthokont subunit 3/11 and opisthokont subunit 16 sequences. We measured the error of ancestral reconstructions as the proportion of ancestral sites that incorrectly contained an indel character (see Supplementary Information, section 6). Plasmids and yeast strains. Bacterial and yeast manipulations were performed using standard laboratory protocols for molecular biology45. Plasmids that were used are listed in Supplementary Information, section 5. Ancestral sequences (pGF140, pGF139, pGF506 and pGF508) were synthesized by GenScript with a yeast codon bias. Triple haemagglutinin epitope tags were included before each stop codon. The Anc.3-11, Anc.16, Anc.3 and Anc.11 genes were subcloned to single-copy, CEN-based yeast vectors. The ADH terminator sequence (247 base pairs (bp)) and Natr drug resistance marker46 were amplified using polymerase￾chain-reaction (PCR) containing 40-bp tails homologous to the 39 end of each coding region and vector sequence. Vectors were gapped, co-transformed into SF838-1Da yeast with PCR fragments and cells were selected for Natr . A second round of in vivo ligation was used to place the ancestral genes under 500 bp of the VMA3 or VMA16 promoters to create pGF140 and pGF139, respectively. For vectors pGF240, pGF241, 1pGF252, pGF253, pGF503–pGF508, pGF510, pGF512–pGF515, pGF517–pGF519, pGF521, pGF523, pGF528, pGF529, pGF531, pGF534–pGF537 and pGF542, the relevant locus (Anc.3-11, Anc.16 or Anc.3) was PCR amplified with 59 and 39 untranslated flanking sequence and cloned into pCR4Blunt-TOPO (Invitrogen). When necessary, a modified Quikchange protocol47 was used to introduce point mutations before the gene was subcloned into a yeast vector (pRS316 or pRS415). To generate pGF502, sequence from codon 31 to the stop codon of Anc.16 was amplified with the ADH::Natr cassette from pGF139, cloned into TOPO, and in vivo ligated down￾stream of the VMA16 promoter (including a start codon) in pRS415. A triple-fragment in vivo ligation was used to generate pGF646–pGF651. Gapped vector containing the VMA16 promoter was transformed into yeast with two PCR fragments of the ring genes to be fused. For pGF646, the coding region of (1) VMA16 (without codons 2–41) and (2) the coding region of Anc.11 (without codons 2–5 were amplified by PCR. The proteolipid on the C-terminal portion of the gene fusion also contained the ADH terminator and Natr cassette; the amp￾lified products contained PCR tails with homology to link the genes to both the gapped vector and to each other. Gene fusions were modelled after the experi￾mental design of Wang et al. (2007)24 in which the lumenal protein sequence linking the two proteolipids was designed to be exactly 14 amino acids. To meet these criteria, additional amino acids were inserted into the following vectors linking the two subunits: pGF646 (Thr-Arg-Val-Asp), pGF648, pGF650 (Thr￾Arg), pGF649, pGF651 (Gly-Ser). Yeast strains that were used are listed in Supplementary Information, section 2. Strains containing deletion cassettes other than KanR 45 were constructed by PCR amplifying the HygR or Natr cassette from pAG32 or pAG25, respectively, with primer tails with homology to flanking sequences to the VMA11 or VMA16 loci. 11D::KanR and 16D::KanR strains (SF838-1Da) were transformed with the HygR and NatR PCR fragments, respectively, and selected for drug resistance. The 11D::HygR locus was amplified and transformed into LGY113 (to create LGY125) and LGY115 (to create LGY124). This was repeated with the 16D::NatR locus to create LGY139 and LGY143. Yeast Growth Assays. Yeast were grown in liquid culture, diluted fivefold and spotted onto YEPD media buffered to pH 5.0 or yeast extract peptone dextrose media containing 25 mM (Figs 2, 3, 4) or 30 mM CaCl2 (Fig. 2f). Whole-cell extract preparation and immunoblotting. Yeast extracts and west￾ern blots were performed as previously described31. Antibodies that were used in this study included monoclonal primary anti-HA (Sigma-Aldrich), anti-Dpm1 (5C5; Invitrogen) and secondary horseradish-conjugated anti-mouse antibody (Jackson ImmunoResearch Laboratory, West Grove, Pennsylvania, USA). Fluorescence microscopy. Staining with quinacrine was performed as previously described31. The cell wall (shown in red) was visualized using concanavalin A tetramethylrhodamine (Invitrogen). Microscopy images were obtained using an Axioplan 2 fluorescence microscope (Carl Zeiss). A 3100 objective, AxioVision software (Carl Zeiss) and Adobe Photoshop Creative Suite (v. 8.0) were used. 32. Lo¨ytynoja, A. & Goldman, N. An algorithm for progressive multiple alignment of sequences with insertions. Proc. Natl Acad. Sci. USA 102, 10557–10562 (2005). 33. Lo¨ytynoja, A. & Goldman, N. Phylogeny-aware gap placement prevents errors in sequence alignment and evolutionary analysis. Science 320, 1632–1635 (2008). 34. Whelan, S. & Goldman, N. A general empirical model of protein evolution derived from multiple protein families using a maximum-likelihood approach. Mol. Biol. Evol. 18, 691–699 (2001). 35. Abascal, F., Zardoya, R. & Posada, D. Prottest: selection of best-fitmodels of protein evolution. Bioinformatics 21, 2104–2105 (2005). 36. Guindon, S. & Gascuel, O. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst. Biol. 52, 696–704 (2003). 37. Anisimova, M. & Gascuel, O. Approximate likelihood-ratio test for branches: A fast, accurate, and powerful alternative. Syst. Biol. 55, 539–552 (2006). 38. Aguinaldo, A. M. A. et al. Evidence for a clade of nematodes, arthropods, and other moulting animals. Nature 387, 489–493 (1997). 39. Yang, Z. PAML 4: Phylogenetic analysis by maximum likelihood. Mol. Biol. Evol. 24, 1586–1591 (2007). 40. Fitch, W. M. Toward defining the course of evolution: minimum change for a specific tree topology. Syst. Zool. 20, 406–416 (1971). 41. Thompson, J. D., Higgins, D. G. & Gibson, T. J. CLUSTALW: improving the sensitivity of progressive multiple sequence alignment through sequence weighting position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22, 4673–4680 (1994). 42. Edgar, R. C. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 32, 1792–1797 (2004). 43. Do, C. B., Mahabhashyam, M. S., Brudno, M. & Batzoglou, S. ProbCons: Probabilistic consistency-based multiple sequence alignment. Genome Res. 15, 330–340 (2005). 44. Fletcher, W. & Yang, Z. Indelible: a flexible simulator of biological sequence evolution. Mol. Biol. Evol. 26, 1879–1888 (2009). 45. Sambrook, J. & Russel, D. W. Molecular Cloning: A Laboratory Manual 3rd edn (Cold Spring Harbor Laboratory Press, 2001). 46. Goldstein, A. L. & McCuster, J. H. Three new dominant drug resistance cassettes for gene disruption in Saccharomyces cerevisiae. Yeast 15, 1541–1553 (1999). 47. Zheng, L., Baumann, U. & Reymond, J. L. An efficient one-step site-directed and site-saturation mutagenesis protocol. Nucleic Acids Res. 32, e115 (2004). LETTER RESEARCH ©2012 Macmillan Publishers Limited. All rights reserved