Un Tour d'Horizon (1/3)

by Christoph
 

When bio­chemists de­ci­phered the ge­netic code in the 1960s (the triplet 'al­pha­bet' for amino acids whose de­fined or­der make up the pro­tein 'words' ) it was – and still is – the most com­pelling evi­­dence for a com­mon "de­scent with mod­i­fi­ca­tions" (Charles Dar­win) of all life on Earth: the al­pha­bet is the same in all life forms (Fig­ure 1). This holds true de­spite some whim­si­cal spelling vari­ants (as we know for every hu­man lan­guage as well ). Not so com­mon are the gram­mar and syn­tax but this should not bother you here (gene ex­pres­sion and its reg­u­la­tion, the how & when to make use of the words, dif­fer con­sid­er­ably among life forms ). Yet un­veil­ing biology's "uni­ver­sal" al­pha­bet came at a hefty price, namely, los­ing the pos­si­bil­ity to trace, un­am­bigu­ously, the evo­lu­ti­o­na­ry his­tory of the 'speak­ers', be they E. coli or ele­phants (Jacques Monod, 1954). If every­body un­der­stand all words, in prin­ci­ple, the de­pen­dence on the parentally in­her­ited vo­ca­bu­la­ry (ver­ti­cal gene trans­mission ) be­comes less of a con­straint. New words can­not only "evolve" from within the in­her­ited vo­cab­u­lary (by mu­ta­tion, do­main swap­ping, and gene du­pli­ca­tion ) but can al­so be picked up from what­ever source is within reach (lat­eral or hor­i­zon­tal gene trans­fer, here ab­breviated HGT ). And, sure enough, lots of words have been ex­changed across all do­mains of life along 3 bil­lion years of evo­lu­tion, and in what­ever form the words came along: ei­ther car­ried by naked "en­vi­ron­men­tal" DNA (set free by lysed cells and dis­rupted viruses ), by viruses and bac­te­rio­phages, by 'gene trans­fer agents' (GTA), or by con­juga­tive plas­mids (think of the alarm­ing spread of an­tibi­otic re­sis­tan­ce ).

Fig­ure 1. Niren­berg & Matthaei doc­u­mented the step­wise elu­ci­da­tion of the ge­netic code on this 15" x 48" chart pieced to­gether from scraps of graph pa­per (Ex­cel spread­sheets weren't yet known in the 1960s ). The 20 ra­dioac­tive ami­no­a­cyl-tRNA prepa­ra­tions are listed across the top, and the 64 RNA triplets are on the ver­ti­cal axis. The pen­cil no­ta­tions in this ma­trix cor­res­pon­ded to amount of la­beled aminoa­cyl-tRNA that bound ri­bo­somes dur­ing ex­pres­sion. Source

We have men­tioned HGT so of­ten in STC that I pick here just three star­tling ex­am­ples: the pea aphid that adopted fun­gal genes to color its skin ei­ther green or pink (Color Me Aphid ), the stick in­sect that adopted bac­te­r­ial genes to bet­ter cope with its ve­gan life style (Picky Sticks), and the case of "porous" ge­netic bar­ri­ers be­tween the bac­te­ria Campy­lobac­ter je­juni and Campy­lobac­ter coli (Species Come and Species Go ). Mol­e­c­u­lar phy­lo­ge­neti­cists soon re­al­ized that any at­tempts to de­rive or­gan­is­mal phy­lo­ge­nies from gene his­to­ries are flawed: there have been way too many HGT events over evo­lu­tion­ary time (that's par­ti­cu­lar­ly true for mi­cro­bial species or lin­eages ). It is still de­bated whether the species con­cept can be re­tained at all for the prokary­otes... (maybe yes, see our re­cent post on the 'species prob­lem' )

Let me turn away from these grandiose views (sens­ing mild hy­poxia al­ready ) and grad­u­ally de­scend from the peaks of the­ory to the lower-ly­ing val­leys and me­an­der­ing rivers of ex­per­i­men­ta­tion. To com­plete this "Tour d'Horizon", we will need to com­plete three stages. In the first stage (this post ), I will show you around a lush val­ley pop­u­lated by a most as­ton­ish­ing col­lec­tion of trees, a plan­ta­tion gen­er­ated by bioin­for­mati­cians who tackle the im­pact of HGT on the larger scale (Fig­ure 2). The sec­ond and third stages of the tour will come later... In the sec­ond stage, I will un­der­take an ar­chae­o­log­i­cal ex­cur­sion along the He­li­case River to show you two of its HGT trib­u­taries. In the fi­nal stage, I will in­vite you to peek into ex­per­i­ments that un­veil as­pects of how HGT ac­tu­ally works on the mol­e­c­u­lar level.

The most as­ton­ish­ing trees

In their 2008 study, Tal Da­gan and her col­leagues asked how sub­stan­tial was the im­pact of HGT on the evo­lu­tion of prokary­otic genomes. They did not la­bo­ri­ously col­lect in­di­ca­tions for HGT by com­paring con­tra­dic­tory phy­lo­ge­netic trees of in­di­vid­ual genes, or gene fam­i­lies. Rather, they chose an al­go­rith­mic ap­proach to com­pre­hen­sively an­a­lyze 181 prokary­otic genomes ‒ 159 Bac­te­ria, 22 Ar­chaea ‒ avail­able back then (to­day, there are >1,500 fully se­quenced prokary­otic genomes ). From the an­no­tated genes (that is, 'open read­ing frames' cod­ing for pro­teins ) of these 181 ge­n­omes they ob­tained a to­tal of 539,723 pro­tein se­quences (the ex­act num­ber is given with­out an er­ror mar­gin; note that even well-cu­rated genomes usu­ally con­tain ~2% mis­an­no­ta­tions ). They then used the 'rec­i­p­ro­cal best BLAST hit ap­proach' (BBH) with a thresh­old of E<1‒10 (for an ex­pla­na­tion, see here; scroll to "What is the Ex­pect (E) value?" ) in a all‑against‑all pro­ce­dure to 'ex­tract' or­thologs. And, by ap­ply­ing the T25 amino acid iden­tity thresh­old (that's roughly cor­re­spond­ing to the limit of what you iden­tify as ho­mologs when scan­ning sequences"by eye" ), they ob­tained 54,349 clus­ters (= pro­tein fam­i­lies ) cov­er­ing 431,492 pro­teins from their pool. They ex­cluded the re­main­ing 108,231 'sin­gle­tons' from fur­ther analy­ses (such 'sin­gle­tons' were likely to en­com­pass most of the mis­an­no­tated genes, so they were on the safe side ). These pro­tein fam­i­lies con­tained the low­est pos­si­ble num­ber of false-pos­i­tive and false-neg­a­tive 'hits' (find some notes of cau­tion here ). Yet, at T25, only six fam­i­lies had mem­bers in all genomes! (Only ) the find­ing that the 181 genomes share at least one gene fam­ily, and there­fore are in­ter­con­nected with each other, there­by form­ing a 'com­plete net­work', al­lowed for a valid ma­trix-based analy­sis of the edges (I stop with the bioin­for­mat­ics lingo right here ).

Fig­ure 2. Three-di­men­sional rep­re­sen­ta­tion of a cir­cu­lar phy­lo­gram for 181 bac­te­r­ial genomes; edges in the up­per hor­i­zon­tal 'plane': an­a­lyzed genomes; ver­ti­cal 'fun­nel': top-down time scale for di­ver­gence. A ref­er­ence tree based on the rRNA genes (5S, 16S, and 23S) of 159 bac­te­r­ial and 22 ar­chaeal genomes/species (grey). B links be­tween dis­tantly re­lated groups that share a gene (green). These links rep­re­sent an­cient HGTs. C links be­tween genes in a tuft of bran­ches and un­re­lated sin­gle ex­tant species (blue). These trans­fers prob­a­bly oc­curred more re­cent­ly, af­ter di­ver­gence of the an­ces­tors of the 181 species. D over­lay of A ‒ C with links bet­ween pairs of species that share genes found in none of the oth­ers (red). These are prob­a­bly the most re­cent HGT events. Source (graph­ics, mod­i­fied); Source (orig­i­nal data)

None of their net­works had more than a shady re­sem­blance to known tax­o­nomic bound­aries, but Da­gan et al. read­ily found three dis­tinct groups with shared cha­rac­ter­is­tics among these net­works. To bet­ter vi­su­al­ize these groups, they pro­jected them onto a ref­er­ence tree re­pre­sen­ting the phy­logeny of the con­cate­nated three rRNAs of all the genomes (Fig­ure 2 A) (most phy­lo­ge­netic stud­ies em­ploy the '16S rRNA ap­proach', which is gen­er­ally con­sid­ered the "gold stan­dard" but is known to fail in re­solv­ing closely re­lated species and the deep­est bran­ching ). In group 1, which they con­sider to rep­re­sent the most re­cent HGTs, links ex­ist be­tween ex­tant genomes, i.e., they share genes that are not present in any of the oth­ers (Fig­ure 2 D, red); In group 2, which likely re­pre­sents 'more re­cent' HGTs (from an an­ces­tor of one branch to an ex­tant genome ), links ex­ist be­tween slightly 'deeper' ed­ges within a bunch of branches and sin­gle ex­tant genomes (Fig­ure 2 C); in group 3, links be­tween ed­ges of branches in dis­tantly re­lated groups rep­re­sent an­cient HGTs (Fig­ure 2 B). Fig­ure 2 shows qual­i­ta­tively (but more im­pres­sive than any ta­ble with dis­crete num­bers could ) that the three groups are equally "dense," that is, HGT was a con­stantly strong con­trib­u­tor to genome evo­lu­tion over time. The au­thors es­ti­mate "that, on av­er­age, at least 81±15% of the genes in each genome stud­ied were in­volved in lat­eral gene trans­fer at some point in their his­tory".

One of their es­pe­cially in­ter­est­ing find­ings was that there are no ap­par­ent "hubs" for HGT, that is, edges in lin­eages that gave rise to mul­ti­ple HGTs over short times. This sug­gests that HGT oc­curs largely at ran­dom and does not de­pend on par­tic­u­larly 'HGT-prone' ge­nomic con­stel­la­tions (in nei­ther 'donor' nor 're­cip­i­ent' ). Also in­ter­est­ing is their find­ing that once a gene was ac­quired by a ge­nome, it tends to stick there and doesn't con­tinue with a 'no­madic lifestyle' (newly ac­quired genes that were sub­se­quently lost within a few mil­lion years would prob­a­bly have slipped through their fil­ter, though  ).

Fig­ure 2 pro­vokes a few com­ments. First, the ver­ti­cal scale for the fun­nel-shaped trees is "years since the emer­gence of dis­tinct rRNA lin­eages", thus roughly 3 bil­lion years, com­pressed to the few mil­lime­ters in the fig­ure. This is as if you look at a map of the United States at a scale that shows San Diego, CA and Boston, MA on a let­ter page (dis­tance ~4,900 km when dri­ving by I‑80 + I‑90 ). You wouldn't be able to put a fin­ger­tip on ei­ther of the two in­ter­sec­tions of in­ter­states 80 and 35 in Des Moines (~10 km apart, roughly equal to 2 mil­lion years ), your fin­ger­tip would in­stead cover a great part of Iowa. We have to ac­cept that time scales cov­er­ing 3 bil­lion years are be­yond hu­man per­cep­tion, and also be­yond the pos­si­bil­i­ties of more gran­u­lar graph­i­cal rep­re­sen­ta­tions. Sec­ond, the dis­tances of the outer edges in the hor­i­zon­tal plane (the in­di­vid­ual genomes, Fig. 2 A ) cor­res­pond to the rel­a­tive sim­i­lar­i­ties of their rRNA se­quences as­sum­ing a con­stant mu­ta­tion rate over time (cer­tainly an over-sim­pli­fi­ca­tion, but we sim­ply don't know bet­ter ) with the time com­po­nent stretched-out ver­ti­cally. This is an in­ge­nious way to "open-up" the vi­sual space for show­ing the dif­ferent 'lay­ers' of HGT events over time. But we are left, as view­ers, with a low-res­o­lu­tion 'gen­eral per­spec­tive' and not al­lowed to zoom-in, or high­light the path(s) of in­di­vid­ual pro­tein fam­i­lies, or to in­ter­ac­tively "climb" within the trees for closer in­spec­tion (which would be im­mensely tempt­ing! ).

Da­gan et al. con­clude that "when all genes and genomes are con­sid­ered, the tree par­a­digm fits only a small mi­nor­ity of the genome at best; hence, more re­al­is­tic com­pu­ta­tional mod­els for the mi­cro­bial evo­lu­tion­ary process are needed." If we con­sider 'all genes and genomes'  here, we're un­for­tu­nately left with the am­bi­gu­ity of whether 'the tree par­a­digm fits only a small mi­nor­ity of the genome', or '…of the genomes'. As­sum­ing that this was just a typo, I could eas­ily fol­low Da­gan et al. in their pro­posal that 'more re­al­is­tic com­pu­ta­tional mod­els' are needed for 'for the mi­cro­bial evo­lu­tion­ary process'. But only if such mod­els could ac­count for the undis­puted fact that the vast ma­jor­ity of cells di­vide by bi­nary fis­sion, which adds a strong tree-like branch­ing com­po­nent to the 'the mi­cro­bial evo­lu­tion­ary process'. In­de­pen­dent of the ne­go­ti­a­tions that are con­ducted in the cells' genomes in be­tween di­vi­sions. The sub­se­quent part of their con­clu­sion, how­ever, will haunt tax­on­o­mist for years to come: "By ac­count­ing for all genes, in­clud­ing the many that are patchily dis­trib­uted across broad tax­o­nomic bound­aries, net­works un­cover a view of mi­cro­bial genome evo­lu­tion (sic!) that in­cor­po­rates HGT as a quan­ti­ta­tively im­por­tant mech­a­nism of nat­ural vari­a­tion among prokary­otic genomes. In con­trast to trees, net­works thus present a means of re­con­struc­ting mi­cro­bial genome evo­lu­tion that ac­com­mo­dates the in­cor­po­ra­tion of for­eign genes…".

 

Other Posts