Commentary In an immunological twilight zone

Early in their evolution, perhaps during their transition from Agnatha (jawless fish) to Gnathostomata (jawed animals), vertebrates committed as much as 1% of their proteinencoding genome to a new system of defense against parasites (1). Central to this anticipatory (adaptive) immune system (2) are three types of antigen receptor—immunoglobulin (Ig), T cell receptor (Tcr), and major histocompatibility complex (Mhc) molecules (3)—each consisting of an antigen-binding part and a part concerned with other functions. In the Tcr and Ig molecules, both parts, the variable (V) and constant (C) domains, are drawn from the multifarious Ig superfamily of proteins; in the Mhc molecules, only the part that does not bind antigen comes from the Ig superfamily, whereas the antigenbinding portion has been derived from a different, as yet unidentified source (4). Although jawed vertebrates represent only a tiny fraction of the living world diversity, the parochialism of our species has led immunologists to devote a disproportionate amount of effort to the study of the anticipatory immune system and to neglect the nonanticipatory system on which all other living forms depend. The seemingly sudden appearance of the anticipatory system and specifically of the three types of antigen receptor is acknowledged by both immunologists and evolutionary biologists as a puzzle worthy of resolution. Among the many questions this puzzle raises, two in particular are fundamental; both questions concern the manner in which the antigen receptors function. The individual Tcr and Ig molecules are quite fastidious in their interactions with antigens, each receptor binding a narrow range of antigens and different receptors binding different antigens. The diversity underlaying this receptor selectivity is generated to a large degree somatically by pasting together various combinations of gene segments and then, often but not always, mutating the resulting pastiche. The first fundamental question is therefore: How did the diversity-generating mechanisms come into being? The Mhc molecules, by contrast, are quite promiscuous in their propensity for antigens and correspondingly lack a somatic diversification mechanism; they possess, however, a different, equally bizarre characteristic: they bind antigens only to be seen in their company by the Tcr (i.e., they function as receptors that ‘‘present’’ antigens to other receptors). Hence the second fundamental question is: When, how, and why has this Mhc restriction of antigen recognition arisen? To answer these two questions, it would be of great help to know how the three receptor types originated. Specifically, evolutionarily minded immunologists have been after ‘‘primitive’’ forms of antigen receptors, forms resembling the common ancestor of Ig and Tcr, assuming there was one. A few years ago Greenberg and his coworkers (5) announced that they may have possibly found one such form. Doubts have persisted, however, whether their ‘‘new or nurse shark antigen receptor’’ (NAR) is really a transitional form between Tcr and Ig or simply an Ig variant. To me, the most recent contribution by this group (6) indicates that the doubts were justified. The nurse shark in which Greenberg and his coworkers (5–7) found the receptor is a representative of cartilaginous fish, the oldest extant branch of jawed vertebrates. The secreted form of the NAR molecule is a homodimer with each chain comprised of six Ig-like domains—one N terminal V-domain and five C-domains. The genes coding for the V-domain undergo somatic diversification in the same way as Tcr and Ig receptors, and their products presumably bind antigen, although this point still needs to be demonstrated formally. So is the NAR an Ig or a Tcr? The authors’ original phylogenetic analysis of the five C-domain protein sequences (5) failed to affiliate this part of the NAR unambiguously with either Ig or Tcr, but subsequent analyses (7–9) indicated a clear relationship to Ig heavy chains. The domains are distinctly related to the Cdomains of the sandbar shark IgW (8) and skate IgX or IgR (10); all three receptors (NAR, IgW, and IgX) are, in turn, related to shark Ig H-chains of the m isotype (9). Hence, as far as the constant part of the molecule is concerned, the NAR is clearly a variant of an Ig molecule—an Ig H-chain isotype. This conclusion is further supported by the observation that like Igs, but unlike Tcrs, the NAR—judging from the presence of corresponding signal sequences (5)—occurs in both soluble and membrane-bound forms. On phylogenetic trees, the V-domain of the NAR seems to be affiliated with the V-domains of some Tcrs (5), and this observation is the sole reason for holding the NAR for a possible intermediate between Ig and Tcr. If, however, the NAR V-domain were really Tcr-like, it would mean either that it has been grafted on the Ig-part of the molecule, for example by an exon shuffling mechanism, or that it acquired its Tcr-likeness by convergent evolution. In the former case, the donor of the grafted domain would presumably be a fully evolved Tcr molecule (gene), whereas in the latter case the origin of the domain would have nothing to do with Tcr. In either case, there would be no reason to consider the NAR a precursor or an intermediate form between Tcr and Ig receptors. The third, and in my mind the most likely possibility is that the Tcr-likeness of the NAR V-domain is an artifact of the phylogenetic analysis. Evolutionary biologists refer to sequence identities below 25% as the ‘‘twilight zone’’ (11), an area of sequence comparisons in which phylogenetic relationships among taxa are extremely difficult, if not impossible to decipher. The similarity of the NAR V-domain sequence to Tcr lies well within the twilight zone. The effect of the twilight zone is illustrated best by comparing the phylogenetic trees in the two successive publications of Greenberg and his coworkers (5–7): The additions of new sequences have changed the clustering of some of the sequences considerably. The available sequence data are simply not good enough to resolve the branching order and the affinities between individual branches in this case. A