Structural convergence and horizontal transfer?

PROTEIN MODULES

CARO-BETH STEWART

Structural convergence and horizontal transfer? The fibronectin evolved

type III module

an immunoglobulin-like

from animals to soil bacteria

appears to have convergently

fold and has probably relatively

Fibronectin is a large, multifunctional protein found in the extracellular matrix and the serum of animals [ 11; it is one of several eukaryotic proteins that appear to have evolved by exon shuffling and duplication of a limited number of structural units, or modules [2-51. Fibronectin is composed of repeats of three different kinds of srructural units, referred to as types I, II, and El [ 11. In a notable recent paper [6], the three-dimensional structure of a fibronectin type III module (Fn3) has been determined and compared to other protein modules of similar structure (Fig. 1). The Fn3 module has a fold that is roughly like those of immunoglobulin C domains 131; it is composed of seven ~-strands, which form two packed antiparallel P-sheets. However, the topology of the Fn3 module is not identical to the immunoglobulin fold; Fn3 is more similar in topology to domain 2 of the T-cell glycoprotein, CD4 (CD4D2), and to domain 2 of the bacterial chaperone protein, PapD (PapDD2) (Fig. 1). Despite the similar topological arrangement of their p-strands, the tertiary structures of Fn3, CD4-D2 and PapD-D2 do not superimpose well in three-dimensional space [6] ; neither are the ammo-acid sequences of Fn3 molecules significantly

recently

been transferred

in evolutionary

time.

similar to those of either PapD-D2 or CD4-D2 [6], These observations led Campbell and co-workers [6,7]to suggest that these three protein modules are probably &e products of convergent evolution to the same generai structure, rather than divergent evolution from a COmfnen ancestral sequence. m evolutionary biology, the term ‘related’implies genetic descent from a common ancestor; in chemistry, this term usually implies only structural or functional similariq, Structural and/or functional similarity of biological macros molecules can be the result of genetic descent from a common ancestor (homology) or independent convergent evolution from genetically unrelated ancestors (analogy). These distinctions are not trivial, as convergence and divergence imply profoundly different things regarding the possibilities for protein folding. If any given protein fold has been re-invented independently numerous times during the evolution of life on earth, then there may be a limited number of possible protein structures. If evolutionarily related, but vastly divergent, sequences can specify that fold, ’then the structure may be more of

(b)

(a)

PapD-IX

Fig. 1. The Fn3 module compared to structures with similar folds [EA.(a) Topolopical arrangement of immunoglobulin C domains, with the o-strands depicted as arrows. The additional P-strands found in V domains are shown attached by dashed lines. (b) Topological arrangement of P-strands of the Fn3, CD4-D2 and PapD-D2 domains. (c) Structures of Fn3, CD4-D2 and PapD-D2 domains, demonstrating their markedly different global folds despite their similar topologies (MOLSCRIPT diagrams, kindly provided by A Main and 1Campbell).

158

@

Current Biology 1993,

Vol

3 No 3

159

an historical accident than a physico-chemical necessity. Either way, cases such as the one discussed above [6] indicate that vastly different primary structures can specify similar three-dimensional structures. Indeed, there are no known cases where structural similarity of evolutionarilyunrelated molecules has arisen by sign&cant sequence convergence. FnMke modules seem to be widespread among modern animal proteins, as judged by similarity to fibronectin type IX ammo-acid sequences; these modules are characterized by a consensus sequence of approximately 90 ammoacid residues [7,81. Sequences that appear to code for FrNike modules have also been found in some bacterial proteins [SlO]. Finding sequences that appear on both sides of the tree of life is usually taken to imply that the sequence was present in the common ancestor of all life on earth. If this were the case for the Fn3like sequences [9], then one would expect to detect them in eukaryotes other than animals, including plants and fungi. However, Peer Bork and Russell Doolittle [8] have made an unusual observation. After extensive scanning of the available genetic databases with consensus motifs derived from presumed Fn3-like modules, they have not detected Fn3 sequences in plants, fungi, or lower eukaryotes, yet they have found several previously unreported occurrences of these sequences in bacterial enzymes. Interestingly, all seven of the known bacterial proteins containing FnMike sequences are carbohydrate-cleaving enzymes from soil bacteria [S] ; six of these enzymes are unrelated. Homologous enzymes in other bacterial species lack the Fn3-like sequences, suggesting that the module is being transferred between bacterial species and genes [8]. The bacterial Fn3 sequences are relatively similar to each other, and unexpectedly similar to some of the animal sequences [8]. The extensive ammoacid sequence similarities between known Fn3 moduleforming sequences [6] and the presumed bacterial Fn3 module-forming sequences [8] (Fig.21 make it unlikely that they are the result of convergent evolution. Careful consideration of these observations led Bork and Doolittie [8] to propose that the Fn3-like module found in bacterial proteins was acquired by horizontal gene transfer from animals.

-Fn3-10 Fn3-9

VSDVPRD GLDSPTG

',;;:y

;;op$;~

CHBl AMEl CLCF DPAl GLCl AMLl

TAPSVP*GN EPATTPKN QAPSVPSG QAGSAPTG :%pPG:

A

B

C

~,_ ..~.,.“ II__., xI‘bc,&z;, .J,,”.~ APAV APRA PSSS PPES

TVR TIT PVT AVT

ASTDN VGVT QSDGY VV ASTDN VGVT AVAN AS ASTDAGSGVA PSADD VAIF

The horizontal transfer of genetic material across species boundaries is an extremely rare event in higher eukaryotes but may be more common in prokaryotes [11,12], especially soil bacteria [ 131. In all organisms, the predominant mode of transmission of genetic material is vertical, from parent to offspring. The vertical nature of genetic transfer allows evolutionary trees to be constructed from molecular sequences which reflect the genealogical history of life on earth. A rigorous way of detecting horizontal gene transfer is to look for a striking discordance between the phylogeny for a set of homologous sequences and the well-established phylogeny of the species from which the sequences were obtained 113,141) which cannot be explained by other mechanisms such as gene duplication, aberrations in evolutionary rates or sequence convergence [ 11,14,15]. If the bacterial Fn3 sequence(s) were horizontally transferred from an animal, then the bacterial sequences should branch with the donor animal lineage, and vice versa. To infer the direction of gene transfer, Bork and Doolittle [8] built evolutionary trees from 39 Fn3-like sequences found in animals and bacteria using two methods, a distance matrix program and a character-state program. Trees built by both programs seem to imply that the bacterial Fn3 sequences are more closely related to the Fn3 sequences found in muscle proteins than are the other animal Fn3 sequences (Fig. 3a), suggesting that the proposed transfer was from animals to bacteria. However, as no Fn.Wke sequences were found in any of the earlier eukaryotic lineages [S], discordance between the expected and observed Fn3 phylogenies is di@ult to document convincingly. Inferences regarding direction of transfer are critically dependent on the placement of the ancestral node, or ‘root’, of the tree. The approach chosen by Bork and Doolittle to root the Fn3 trees was by ‘midpoint’: in this method the root is placed ha.lf%ayalong the longest lineage reconstructed between two sequences by the tree building program. For the midpoint method to produce an accurately rooted tree, the ditferent molecular lineages must have evolved at approximately the same rate. Although calculations indicate that Fn3 sequences found in different proteins are evolving at roughly equal rates

--

C’

E

,:i&y :;>y;\Tp$ l_“ (II. r*

GGNSFVQEFTVPGSK!$&%$$GLKPG HFSGRPREDRVPHSRt@,X$'@tTPG NGPGPTKTKTAGPDQ@&%@GLQPT NLPGEHGQRLPISRNT@~i$GLSPG '~;>>