Ac­ci­den­tal Com­plex­ity

by Christoph

…one would have to say that nat­ural se­lec­tion does not work as an en­gi­neer works. It works like a tin­kerer – a tin­kerer who does not know ex­actly what he is go­ing to pro­duce but uses what­ever he finds around him…to pro­duce some kind of work­able ob­ject.
—François Ja­cob, 1977 (1)

You may won­der why you read the same motto twice in a week in STC but there's a rea­son. I will fol­low up Merry's re­cent post, her lu­cid de­scrip­tion of in sil­ico and in vivo ex­per­i­ments demon­strat­i­ng that evo­lu­tion can achieve com­plex­ity by small in­cre­men­tal steps if they con­fer some mar­ginal ben­e­fit. No "big leaps" are re­quired, no "mas­ter plan" – tin­ker­ing is suf­fi­cient. If you have some ex­perience with DIY projects you know you can tin­ker all your way to a so­lu­tion that "works just fine", given you're pa­tient enough. In some cases you prob­a­bly re­al­ize af­ter­wards that the so­lu­tion you found is way more com­pli­cated – or even more com­plex – than nec­es­sary. But you keep and use your overly com­plex DIY so­lu­tion be­cause it, well, works just fine. Thus, if evo­lu­tion were a tin­kerer rather than an en­gi­neer – as François Ja­cob pro­posed, and for which Merry dis­cussed avail­able evi­dence – bi­ol­o­gists should also find ex­am­ples for such un­nec­es­sar­ily com­plex so­lu­tions that oc­cur­red... by ac­ci­dent. So­lu­tions that would in­crease the com­plex­ity of a sys­tem with­out pro­vid­ing any ap­par­ent ben­e­fit that could be se­lected for.

The case of the fun­gal V‑ATPases

Among the most highly con­served en­zymes through­out all of life are the multi-sub­unit AT­Pases of the F‑, V‑, and A‑types. These en­zymes cou­ple the trans­port of H⁺ or small cations (Na⁺, K⁺, Ca²⁺) across a mem­brane to the syn­the­sis and/or hy­drol­y­sis of ATP. They share a sim­i­lar dy­namo-like ar­chitecture with a mem­brane-em­bed­ded do­main ('ro­tor') that trans­ports ions across this mem­brane, and a cy­toso­lic do­main ('sta­tor') that, de­pend­ing on the di­rec­tion of ro­ta­tion of the ro­tor do­main, ei­ther hy­drolyzes or syn­the­sizes ATP. Ear­lier, we fea­tured the bac­te­r­ial F1Fo AT­Pase (ATP syn­thase) in our Pic­tures Con­sid­ered se­ries. The eu­karyal vac­uo­lar-type H+-ATPases (V‑ATPases) pump H⁺ ions across in­tra­cel­lu­lar and plasma mem­branes dri­ven by the hy­drol­y­sis of ATP and acid­ify vac­uoles (among other func­tions). The Vo do­main ('ro­tor') of the eu­karyal V‑ATPases are ring-shaped pro­tein mul­ti­mers – 'hetero-hexamer's in pre­cise ter­mi­nol­ogy – of five Vma3 sub­units and one Vma16 sub­unit. What is true for An­i­mals and Plants is not nec­es­sar­ily also true for the Fungi: the Vo do­main of those Fungi for which se­quence data are avail­able is as­sem­bled from four Vma3 sub­units, one Vma11 and one Vma16 sub­unit (Fig­ure 1a). This ap­par­ent in­crease in com­plex­ity of the fun­gal ver­sion of the V‑ATPase prompted Joe Thorn­ton and his cowork­ers, then at the Uni­ver­sity of Ore­gon, to study the Vo do­main of Sac­cha­romyces cere­visiae  in more de­tail (2).

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 Vo ring is mem­brane bound. Pro­tein sub­units are la­beled and col­ored: Vma3 (blue), Vma11 (yel­low), and Vma16 (red). b Max­i­mum like­li­hood phy­lo­ge­netic tree 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. Col­ors cor­re­spond to those of sub­units in panel a; undu­pli­cated or­thologs of Vma3 and Vma11 are green. As­ter­isks show ap­proximate like­lihood ra­tios for ma­jor nodes: ****, >10³; ***, >10²; **, >10; *,<10; ~, <2. Source

They first in­ferred the phy­logeny of the Vma3, Vma11, and Vma16  pro­tein fam­ily us­ing the se­quences from 139 ex­tant species (Fig­ure 1b). Their phy­logeny in­di­cates that Vma3 and Vma11 are par­alogs, that is, sis­ter pro­teins. They are, in all like­li­hood, de­scen­dants of an an­ces­tral pro­tein (in Fig. 1b: Anc.3–11) whose gene was du­pli­cated ~800 mil­lion years ago in the last com­mon an­ces­tor of all Fungi. It seems likely that the Vma3/Vma11 and Vma16 lin­eages also de­scend from an older gene du­pli­ca­tion deep in the eu­kary­otic lin­eage. But due to the scarcity of se­quence data this was presently im­pos­si­ble to prove as no mean­ing­ful 'out­group' could be in­cluded in this phy­lo­ge­netic analy­sis.

I should men­tion here that gene du­pli­ca­tions are very com­mon in all do­mains of Life. For ex­am­ple, about 30% of the genes of the E. coli may have their ori­gin in dup­li­cations. Gene du­pli­ca­tions arise from 'hic­cups' of the rep­li­cation ma­chin­ery that are not im­me­di­ately re­paired but be­come fixed in a genome if they con­fer some se­lec­tive ad­van­tage. One ex­am­ple was given here in STC: an an­ces­tral fun­gal gene cod­ing for a trans­porter of ba­sic ami­no acids was ac­ci­den­tally du­pli­cated and both copies kept be­cause one copy, LYP1, dif­fer­en­ti­ated fur­ther into a ly­sine per­me­ase while the other, ALP1, dif­fe­ren­ti­a­ted into an argi­nine per­me­ase.

Finni­gan et al. took their ap­proach one step fur­ther and 'guessti­mated' by a max­i­mum like­li­hood al­go­rithm the an­ces­tral amino acid se­quences with the high­est prob­a­bil­ity of pro­duc­ing the ex­tant se­quences (2). This, in turn, al­lowed them to re­verse en­gi­neer the an­ces­tral pro­teins Anc.3–11 and Anc.16 that should have made up the an­cient two-par­a­log eu­kary­otic ring of the Vo sub­com­plex, and, in ad­di­tion, the du­pli­cated sub­units Anc.3 and Anc.11 from the three-com­po­nent ring in the com­mon an­ces­tor of all Fungi (Fig­ure 1b). They were then ready to put these in sil­ico re­sults un­der scrutiny in vivo.

Res­ur­rec­tion in vivo

S. cere­visiae strains lack­ing com­po­nents of the Vo ring of their V‑ATPase re­main vi­able but loose the abil­ity to grow at el­e­vated Ca²⁺ concentra­tion. This al­lowed Joe Thorn­ton and his cowork­ers to test their bona fide 'an­ces­tral' pro­teins in a com­par­a­tively sim­ple com­ple­men­ta­tion as­say by in­tro­duc­ing plas­mid borne an­ces­tral genes into rec­i­p­ro­cally de­fi­cient yeast cells. Func­tion­al­ity of the in­troduced genes would re­sult in restora­tion of growth of the re­cip­i­ent yeast cells at el­e­vated Ca²⁺ concen­tra­­tions. They found that Anc.3–11 res­cues the growth at el­e­vated Ca²⁺ concentra­tions of ei­ther vma3Δ or vma11Δ yeast mu­tants, in­di­cat­ing that the func­tions (= oligomer­iza­tion pro­per­ties) of the present-day ver­sions of Vma3 and Vma11 were al­ready present be­fore the du­pli­ca­tion that gen­er­ated them. In ad­di­tion, Anc.3–11 could par­tially res­cue growth of a vma3Δ­vma11Δ dou­ble mu­tant, which nei­ther of the present-day ver­sions Vma3 and Vma11 could. The re­con­structed Anc.16 res­cued growth in a vma16Δ mu­tant, and co-ex­pres­sion of Anc.3–11 and Anc.16 to­gether res­cued cell growth in a vma3Δvma11Δvma16Δ triple mu­tant lack­ing all three ring sub­units. These re­sults es­tab­lished that dur­ing the in­crease in com­plex­ity, nei­ther the Vo com­plex nor its com­po­nent pro­teins evolved new func­tions re­quired for growth un­der the con­di­tions in which the ring is known to be im­por­tant.

By con­tin­u­ing this 're­com­bi­na­to­r­ial game' for the three-com­po­nent ring sys­tem, they found that Anc.3 res­cues growth of a vma3Δ mu­tant but not of a vma11Δ mu­tant, while, rec­i­p­ro­cally, Anc.11 re­stores growth of a vma11Δ but not of a vma3Δ mu­tant. And, fi­nally, co­ex­pres­sion of Anc.3 and A.11 was nec­es­sary to res­cue growth of a vma3Δvma11Δ dou­ble mu­tant. These re­sults in­di­cated that af­ter their ori­gin by gene du­pli­ca­tion, Anc.11 lost the abil­ity of its an­ces­tral pro­tein, Anc.3–11, to carry out at least some func­tions of Vma3, while Anc.3 lost the an­ces­tral ca­pac­ity to carry out those of Vma11.

Fig­ure 2. The an­ces­tral Vo ring of the V‑ATPase com­plex had two types of sub­units (red, green). Black squares, tri­an­gles and cir­cles show inter­faces of the pro­teins that can bind to other sub­units via in­ter­faces marked with cor­re­spond­ing holes. In fungi, one of the genes du­pli­cated (red square), pro­duc­ing two types of sub­units (blue, yel­low). Some of the sub­units lost sites where other sub­units could bind, marked here by red as­ter­isks. (Sub­unit col­ors as in Fig. 1). Source

Ear­lier stud­ies had es­tab­lished the highly or­dered spa­tial arrange­ment of the Vma16, Vma11, and Vma3 sub­units in the Vo ring of the yeast V‑ATPase, con­strained by the ca­pac­ity of each sub­unit to form spe­cific in­ter­faces with the other sub­units (Fig­ure 2 shows schemat­i­cally the sub­unit arrange­ment in a 'lin­earized' ver­sion). Vma11 (yel­low) is re­stricted to a sin­gle po­si­tion be­tween Vma16 (red) and Vma3 (blue), be­cause its right "ar­row hole" in­ter­face can only bind to the "ar­row" in­ter­face on the left side of Vma16 and its left "square" in­ter­face can only bind to the "square hole" in­ter­face on the right side of Vma3. Vma16 is re­strict­ed to a sin­gle po­si­tion be­tween Vma11 (yel­low) and Vma3 (blue), be­cause its "cir­cle hole" in­ter­face on the right side can only bind to the "cir­cle" in­ter­face on the left side of Vma3. Only Vma3 can bind ei­ther to cop­ies of it­self, or to Vma16, or to Vma11. Be­cause Anc.3–11 can com­ple­ment the loss of both Vma3 and Vma11, Finni­gan et al. as­sumed that Anc.3–11 could par­tic­i­pate in all three spe­cific in­ter­ac­tion in­ter­faces, and that these ca­pac­i­ties were then par­ti­tioned be­tween Anc.3 and Anc.11 af­ter the du­pli­ca­tion of Anc.3–11. To test this they cre­ated a set of rec­i­p­ro­cal gene fu­sions and tested them for growth restora­tion in the pres­ence of el­e­vated Ca²⁺ con­cen­tra­tions in the ap­pro­pri­ate yeast mu­tants. They could thus es­tab­lish that sub­unit in­ter­ac­tions oc­cur in the res­ur­rected pro­teins in the same man­ner as in the present-day ver­sions of Vma3 and Vma11. Anc3 and Anc.11 had ap­par­ently both lost each one in­teraction in­ter­face due to mu­ta­tions early af­ter du­pli­ca­tion of the an­ces­tral Anc. 3–11 gene.

In sum­mary, the re­sults ob­tained by Finni­gan et al. make a strong ar­gu­ment for the view that ar­chitectural com­plex­ity of mol­e­c­u­lar as­sem­blies can in­crease when a few mu­ta­tions de­grade an­cest­ral in­ter­faces but leave other func­tions in­tact. The spe­cific roles of the Vma3 and Vma11 sub­units of the Vo sub­com­plex seem to have been ac­quired when the du­pli­cated genes lost some, but not all, of the ca­pac­ity of the an­ces­tral pro­tein Anc.3–11 to par­tic­i­pate in in­ter­ac­tions with copies of it­self and the other pro­tein re­quired for proper ring as­sem­bly, Anc.16. Be­cause com­ple­men­tary losses oc­curred in both lin­eages, the two de­scen­dant sub­units be­came ob­lig­ate com­po­nents, and the com­plex­ity of the ring in­creased. This "com­plex­ity by ac­ci­dent" found for the fun­gal V‑ATPases and, for in­stance, the evo­lu­tion of pre-trans­la­tional RNA edit­ing (not to be con­fused with RNA splic­ing) have led to the con­cept of 'Con­struc­tive Neu­tral Evo­lu­tion' (CNE), which chal­lenges the more clas­si­cal view of se­lec­tion as the sole source of cre­ativ­ity and di­rec­tion, as the only force that can cause trends or build com­plex fea­tures. This de­bate goes on... it's sci­ence, af­ter all!

 

Ref­er­ences

(1) Ja­cob F. 1977. Evo­lu­tion and tin­ker­ing. Sci­ence, 196 (4295), 1161 – 1166. PMID 860134 (Pay­wall; Open Ac­cess PDF here)

(2) Finni­gan GC, Han­son-Smith V, Stevens TH, Thorn­ton JW. 2012. Evo­lu­tion of in­creased com­plex­ity in a mol­e­c­u­lar ma­chine. Na­ture, 481 (7381), 360 – 364. PMID 22230956 (Pay­wall; Open Ac­cess PDF here)