Conservation of amino acid sequences in albumin: is albumin an essential protein?

Albumin, the most abundant protein in vertebrate serum, is thought to be a nonessential protein because other serum proteins take over albumin’s actions to bind and transport hydrophobic ligands such as fatty acids, bilirubin, and steroids and to maintain osmotic pressure in blood in “analbuminemic” humans (Gitlin and Gitlin 1975; Bowman et al. 1976) and rats (Nagase et al. 1979). Additional support for this idea comes from a recent analysis (Minghetti et al. 1985) of the amino acid sequences of rodent and human albumin and alpha-fetoprotein (AFP), a paralogue of albumin (Alexander-Eiferman et al. 198 1; Jagodzinski et al. 198 1; Morinaga et al. 1983). This analysis shows that albumin and AFP are accumulating amino acid changes about twofold and threefold, respectively, faster than does hemoglobin. In fact, the rate of change in AFP is about half that of pseudogenes and approaches that of fibrinopeptides (Minghetti et al. 1985). When segments of a protein undergo relatively rapid changes in their amino acid sequences, they are thought to be nonessential to the biological functioning of the protein. Thus, if most of albumin is undergoing rapid change, and if the absence of albumin does not lead to an obvious disease state, then it would seem that albumin no longer has an essential biological function in the organism. Here, I show that both of these premises are invalid, and I propose that albumin has essential, albeit still unelucidated, biological function(s). First, it has not truly been demonstrated that humans or rats can survive without albumin, because “analbuminemic” humans and rats have lo-25 pg albumin/ml. Even though this concentration is > 1,000 fold less than the normal albumin concentration, it still is high when compared with the concentration of growth factors or other essential proteins in serum. Second, as shown in table 1, exons 1214, the last three coded exons in domain III of human albumin, are 50% identical to human AFP. Figure 1 shows the alignment of the 58 residues of exons 13 and 14 in human albumin and AFP. To put these numbers in perspective, note that alignment of human hemoglobin alpha and beta chains, a paralogous system, reveals 45% identities (Feng and Doolittle 1987). These genes diverged -400-450 Mya (Goodman et al. 1988), a date similar to that for the divergence of albumin and AFP. Thus, a 133-residue part of the human albumin and AFP (exons 1214) is changing more slowly than the alpha and beta chains of human hemoglobin-and clearly much less than fibrinopeptides and pseudogenes. From table 1 it is clear that other subdomains of human albumin and AFP, such as subdomains I-C, II-A, II-C, and II-D, also are under constraints as far as changes in amino acid sequences are concerned. Table 1 reveals that some albumin and AFP subdomains (e.g., I-A, I-D, II-B, and III-A) differ by 65%-77%, which serves to further emphasize that there are indeed constraints on changes in the sequences of other parts of albumin and AFP, parts that I suggest are involved in essential biological actions of albumin and AFP.