Metal Metabolism: Transport, Development and Neurodegeneration

Albumin as a zinc carrier: properties of its high-affinity zinc-binding site

Jin Lu, Alan J. Stewart, Peter J. Sadler, Teresa J.T. Pinheiro, Claudia A. Blindauer


Although details of the molecular mechanisms for the uptake of the essential nutrient zinc into the bloodstream and its subsequent delivery to zinc-requiring organs and cells are poorly understood, it is clear that in vertebrates the majority of plasma zinc (9–14 μM; approx. 75–85%) is bound to serum albumin, constituting part of the so-called exchangeable pool. The binding of metal ions to serum albumins has been the subject of decades of studies, employing a multitude of techniques, but only recently has the identity and putative structure of the major zinc site on albumin been reported. Intriguingly, this site is located at the interface between two domains, and involves two residues from each of domains I and II. Comparisons of X-ray crystal structures of free and fatty-acid bound human serum albumin suggest that zinc binding to this site and fatty acid binding to one of the five major sites may be interdependent. Interactive binding of zinc and long-chain fatty acids to albumin may therefore have physiological implications.

  • albumin
  • fatty acid
  • metal specificity
  • serum
  • zinc trafficking


Serum albumin is, with an average concentration of 600 μM, the most abundant protein in the blood plasma of vertebrates and functions as a carrier for a variety of nutrients, meta-bolites and xenobiotics, as reviewed previously [1,2]. The present paper briefly reviews the metal-binding properties of albumin in general, before highlighting recent advances in understanding zinc binding to albumin.

Albumin is thought to be the major zinc transporter in plasma [3], and typically binds approx. 80% of all plasma zinc [4]. Numerous studies have also demonstrated that albumin modulates zinc uptake into cells [5,6], and some suggest that, at least in some cell types, such as endothelial cells, there is a receptor-mediated endocytosis pathway for albumin-bound zinc uptake [7], and that zinc transport across endothelia involves co-transport with albumin [8]. Furthermore, certain cases of familial hyperzincaemia appear to be due to increased zinc binding to albumin [9].

In the light of the global importance of an adequate zinc supply for many physiological processes, including cell division and differentiation, combined with considerable research activity on albumin, it is therefore surprising that the major zinc site on HSA (human serum albumin) has been identified only relatively recently [10].

Metal binding to albumins

The metal-binding capacity of albumins has been acknowledged for a long time [11], and albumin has been shown to bind a variety of essential and toxic metal ions, including Ca(II), Co(II), Ni(II), Cu(II), Zn(II) and Cd(II). Many studies have addressed questions of affinity, stoichiometry and specificity [1220].

Although there is still a considerable lack of detailed structural information, four different metal sites have been described: (i) the so-called ATCUN (N-terminal copper and nickel-binding) motif at the N-terminus, the primary site for Cu(II) and Ni(II) [21]; (ii) the so-called site A, the primary Zn(II)-binding site that can also bind other 2+ metal ions [22]; (iii) site B, which displays a high affinity towards Cd(II), but the location of which is unknown [12,14,18]; and (iv) the reduced thiol of Cys34, which binds gold [23] and platinum [24] compounds. Although this site has repeatedly been suggested to also bind Cd(II), we have demonstrated that this is not the case by studying both Cys34-blocked BSA [18] and the C34A mutant of HSA [10] by 113Cd or 111Cd NMR spectroscopy. Finally, although Ca(II) is thought to be transported by albumin in the plasma [25], the identity of its binding site(s) is unknown, although the so far unidentified site B may play a role [18].

The ATCUN motif is probably the most studied metal-binding site on albumin (for a comprehensive review, see [26]). Such sites are not restricted to albumin, and require the presence of a histidine residue in position 3 of an amino acid chain. The ATCUN motif is particularly suitable for metal ions that have a prevalence for tetragonal or square-planar co-ordination, and provides in total four nitrogen ligands that include an imidazole nitrogen of His3, the N-terminal amino group and two deprotonated backbone amide nitrogens. In addition, the carboxylate side chain of Asp1 has also been implicated in metal binding, giving an overall penta-co-ordinate site [26], although this residue appears to be not crucial for high-affinity Cu(II) and Ni(II) binding. The X-ray structures of copper complexes of corresponding tripeptides confirm that Cu(II) is co-ordinated in a mildly distorted square-planar fashion, with very weak interaction to a fifth apical ligand [27]. In principle, as shown in studies involving tripeptides [28], other metal ions, such as Zn(II) and Co(II), can also bind to the ATCUN motif, and have been shown to effect backbone amide deprotonation at physiological pH, but to a lesser extent than does Cu(II). Interestingly, dog, pig and chicken albumins are devoid of His3, and this has been correlated with the higher susceptibility of dogs to copper toxicity [17]. Typical affinities for Cu(II) for the ATCUN site are of the order of log K≈11 [17] for the intrinsic stoichiometric stability constant.

A detailed comparative study of HSA and dog serum albumin [14] has demonstrated further that neither Zn(II) nor Cd(II) binds preferentially to the N-terminus, and a recent study suggests that neither does Co(II) [29], although this is controversial [30,31]. The equilibrium dialysis experiments reported by Goumakos et al. [14] also revealed that, in total, 3 mol equiv. of Zn(II) and Cd(II) can be bound to albumin with non-negligible affinity, but that there is only one high-affinity site for Zn(II) and two for Cd(II) on both HSA and dog serum albumin. The apparent stability constants for the Zn(II) site on HSA was determined as log K=6.4±0.8, and that for the two Cd(II) sites as log K=5.3±0.6. Intrinsic stoichiometric constants for Zn(II)–HSA have been reported as log K=7.53±0.45 [16] and log K=7.1±0.2 [19]. It should be noted that these constants are independent of the protonation state of albumin, but still depend on factors such as temperature and ionic strength, which may account for discrepancies between the results of experiments from different laboratories.

Clues to the nature of the preferred binding site for Zn(II) came mainly from 113Cd-NMR studies [12,18]. BSA [12] as well as HSA and albumins from sheep, pigs and horses [18] all display two strong binding sites for Cd(II). The corresponding 113Cd spectra contain two peaks, one at 25–30 p.p.m. (labelled B), and another at 110–150 p.p.m. (labelled A), and both peaks are rather broad (linewidths 200–900 Hz). The 113Cd chemical shifts observed for the two sites are characteristic of mixed nitrogen/oxygen ligand sets [32], and site A was suggested to contain no fewer than two nitrogen ligands [12,14], whereas site B is thought to contain one or no nitrogen ligand.

One of the most crucial conclusions arising from competition experiments was the pronounced preference for Zn(II) of site A, as 1 mol equiv. is sufficient to completely suppress the corresponding 113Cd or 111Cd resonance in either BSA [12] or HSA [10,18] (Figure 1), suggesting replacement of Cd(II) by Zn(II). In contrast, higher ratios of either Cu(II) and Ni(II) are required to affect peak A, with Ni(II) displaying the smallest effect. This is consistent with preferential binding of Cu(II) and Ni(II) to the N-terminus as discussed above, and also indicates that site A might not be particularly suitable for accommodating the d8 ion Ni(II).

Figure 1 111Cd NMR spectroscopy of recombinant HSAs in the presence of 2 mol equiv. of 111Cd(II)

Albumin defatted either through dialysis or charcoal treatment shows binding of 111Cd(II) to two different sites. Addition of 1 mol equiv. (eq) of Zn(II) leads to complete suppression of peak A, presumably by displacement of 111Cd(II). Mutation of the putative zinc ligand His67 to alanine also impairs Cd(II) biding to this site. Saturation of wild-type albumin with fatty acid (+5 eq octanoate) also interferes with Cd(II) binding to site A. The preferred Cd(II)-binding site B is not affected by any of these manipulations.

CD and EPR spectroscopy and metal competition experiments [22] also demonstrated that the second mol equiv. of Cu(II) added to BSA or HSA binds to the high-affinity Zn(II) site. The site displays fairly rigid co-ordination geometry and binds Zn(II) more strongly than either Ni(II) or Cd(II).

In summary, these and other competition experiments allowed the following conclusions: (i) there are, in total, three sites with non-negligible affinity for Zn(II), Cd(II) and Cu(II) on both BSA and HSA; (ii) HSA and BSA have exactly one high-affinity site for Zn(II) and Cu(II), and two for Cd(II), one of which is identical with the Zn(II) site; (iii) the high-affinity Zn(II) site also binds other 2+ metal ions, including Cu(II), if the N-terminus is saturated or unavailable as in dog albumin; and (iv) both Zn(II) and Cu(II) have a higher affinity than Cd(II) for this site.

The high-affinity zinc site on albumin

In the light of the importance of zinc transport and delivery, we set out to identify and characterize the high-affinity zinc site on HSA, using a combination of site-directed mutagenesis, molecular modelling and NMR spectroscopy [10]. An inspection of published X-ray crystal structures of HSA revealed that the only region which contains two histidine ligands in close proximity to potential oxygen donors involves His67 and His247 (Figure 2). Mutation of His67 to alanine had a major effect on cadmium binding to site A, as judged by 111Cd NMR of the mutant (Figure 1), suggesting that this residue may indeed be involved in metal binding to this site.

Figure 2 Location and modelled structure of the high-affinity zinc-binding site on albumin [10]

(a) The zinc site is at the interface of domains I (orange) and II (blue). The model is based on PDB code 1AO6 [33]. (b) The zinc (purple) site is composed of two amino acids from domain I (His67 and Asn99), two amino acids from domain II (His247 and Asp249) and a fifth exogenous ligand.

One of the peculiar characteristics of the NMR resonances of site A, as reported from various studies, is the relatively large chemical shift range observed for this site. A close scrutiny of available data revealed that one of the major differences between the various preparations was the composition of the buffers, and our experiments involving titrations with chloride indeed confirmed that the chemical shift of 111Cd(II) in site A is highly dependent on the availability of exogenous ligands [10], suggesting that Cd(II) in this site might bind an additional ligand.

The molecular model of Zn(II)-bound albumin shown in Figure 2 is compatible with all available experimental data. Intriguingly, the site is located at the domain interface between domains I and II, and each domain provides two ligands: His67 and Asn99 from domain I, and His247 and Asp249 from domain II. This site is essentially preformed in all published X-ray crystal structures of fatty-acid-free albumin. Starting from an X-ray crystal structure of fatty-acid-free albumin, only minor adjustments to atom positions are required to model a physically reasonable zinc site: the root mean square deviation between the X-ray crystal structure used as starting model (PDB code 1AO6) [33] and the modelled structure is only 0.73 Å (1 Å=0.1 nm) for all heavy atoms of the zinc ligands [10]. The inclusion of a fifth ligand, modelled as a water molecule, is also supported by our modelling studies, as attempts to model a four-co-ordinate site led to a highly distorted co-ordination geometry with an ‘empty’ space towards the exterior.

Compared with the major metal-binding side chains of cysteine, histidine, aspartate and glutamate, asparagine (as well as glutamine) is a rare ligand for zinc (or other metals) in proteins, but a few X-ray crystal structures of Zn(II)-bound asparagine are available. The active sites of purple acid phosphatase [34], calcineurin [35], periplasmic 5′-endonucleotidase [36] and the inactive form of the glycyl-glycine endopeptidase LytM [37] have the same amino acid set as the Zn(II) site on albumin. Whereas in the former three proteins, the Zn(II) ion is in a bimetallic cluster, the active site of LytM contains a single Zn(II) ion. LytM is an autolysin, and, notably, the latent form in which Zn(II) is tetrahedrally co-ordinated by two histidine, one aspartate and one asparagine residue is devoid of a water ligand, and displays very low enzymatic activity. It is thought that activation involves cleavage of the N-terminus of the protein which contains the Asn117 ligand, yielding a free co-ordination site for the substrate. Thus, by analogy to the ‘cysteine-switch’ in matrix metalloproteases, the activation mechanism for LytM has been termed the ‘asparagine-switch’. In summary, considering the apparent requirement for either a second metal ion or removal of one of the four ligands from a Zn(His)2(Asp)(Asn) site, the Zn(II)-binding site on albumin is not expected to show notable hydrolytic activity.

Finally, sequence comparisons of all available serum albumins show that the four proposed ligands are conserved in almost all available mammalian sequences, except for that of the guinea-pig, which lacks both domain I residues. The zinc site is also present in the marsupial opossum albumin, but other vertebrate albumin sequences from fish, snakes, amphibians and birds also lack one or more of the zinc-binding residues. These observations suggest that it would be interesting to compare mechanisms of zinc transport and metabolism in these different species.

Interactive binding of metal ions and fatty acids

It had been noted that the appearance of peak A in 113Cd NMR spectra is dependent on the fatty acid content of the albumin preparation used [18], pointing towards the possibility for reduced Cd(II) affinity for albumin in the presence of fatty acids, although previous studies had not reported a correlation between the presence of fatty acids and Zn(II) affinity [14,15]. We tested this hypothesis using recombinant HSA [10], and compared the 111Cd NMR spectra of a sample containing approx. 8 mol equiv. of octanoate with that of an extensively dialysed sample (Figure 1). Clearly, the presence of an excess of octanoate had an effect on Cd(II) binding to site A, but had none on site B. A large number of X-ray crystal structures with various bound fatty acids is available (e.g. [38,39]), and a comprehensive analysis of all these structures reveals that, in each one, the Zn(II) site A is disrupted. This disruption is caused by a domain–domain movement that can be described as a rotation of domain I with respect to domain II [38]. As a consequence, the two zinc-binding ‘braces’ in the two domains move apart by 4–6Å [10] (Figure 3).

Figure 3 Schematic illustration of the relative domain I–domain II movement induced by fatty acid binding in HSA and its impact on the zinc site

(a) Residues involved in fatty-acid-binding site FA2 [38] in fatty-acid-free albumin (with modelled zinc site [10]; the zinc ion is coloured purple), with the van der Waals surfaces of domain I residues in orange and those of domain II residues in blue. (b) The same residues in albumin containing five molecules of myristate (PDB code 1BJ5). The straight lines are drawn to illustrate the rotation of domain I with respect to domain II, and represent arbitrary axes through the middles of the pocket-forming residues in subdomains IA (the larger portion) and IB (the small portion on the lower right). The small arrows indicate the resulting movement of the domain I zinc-binding residues in relation to the domain II residues.

Intriguingly, the fatty acid molecule that is largely responsible for this particular conformational change is, like the zinc ion, bound to a site (FA2) which is formed by two half-sites, one in each of the two domains (Figure 3). The full fatty-acid-binding site is only formed when the conformational switch engages the relevant side chains in both domains (see legend of Figure 3). Thus binding of long-chain fatty acids to FA2 and zinc binding to site A might be mutually exclusive.

The observations described above raise the question of the significance of the allosteric relationship between zinc and fatty acid binding to albumins. HSA contains both high- and low-affinity fatty-acid-binding sites [40], and there is agreement about the existence of two high-affinity sites for medium- and long-chain fatty acids, followed by three to four sites with only slightly lower affinities [38,41]. It is also clear that chain length plays a major role [41,42]. The most common fatty acids bound to HSA in vivo are the longchain fatty acids oleate and palmitate [2], and, under normal healthy conditions, each albumin molecule in the human plasma carries one or two fatty acid molecules in their deprotonated (carboxylate) forms.

The average occupancy of the zinc site on albumin is much lower, with only approx. 2% of circulating albumin molecules thought to carry a zinc ion [2]. Also, the apparent binding constants for the first two equivalents of long-chain fatty acids are larger than that of zinc [41]. Taking these considerations together, it seems unlikely that zinc binding would influence the amount of fatty acid transported under normal conditions. However, we hypothesize that the zinc-binding affinity of albumin might be reduced during conditions of increased free fatty acid mobilization, e.g. after strenuous exercise, when about four molecules of fatty acid may be bound to albumin [2]. Changes in plasma levels of zinc during exercise have been reported (e.g. [43]), although there does not appear to be clear agreement as to whether these levels are increased or decreased.


Albumin contains several metal-binding sites with clear specificities for different metal ions. Specificity is achieved by exploiting differences in the co-ordination chemistries of the preferred metal ions. Under physiological conditions, only Zn(II) will be bound to the high-affinity site, as Cu(II) binds much more avidly to the ATCUN motif, and other metal ions, including the toxic Cd(II), have a lower affinity for the zinc site.

Albumin-bound zinc has been categorized as ‘exchangeable’ [11], and this is thought to be important with respect to transport and delivery of zinc. A site that works in the physiological transport of zinc has to have thermodynamic and kinetic properties that are suited to this task. Accordingly, metal ions bound to site A seem to have rapid exchange kinetics, as judged from competition experiments involving UV–visible spectroscopy. The conditional dissociation constant of the zinc albumin complex (approx. 1 μM) is relatively high compared with other zinc–protein complexes, but it is yet about one order of magnitude lower than typical plasma zinc concentrations (approx. 10 μM). Taking into account the approx. 100-fold excess of albumin and the presence of other zinc-binding ligands in plasma, this is exactly sufficient to preclude the existence of significant amounts of ‘free’ Zn(II) which could be harmful. Furthermore, there are indications that the physiological state of the organism influences the affinity of albumin for zinc via the binding of excess fatty acids. This might have functional significance, but further studies are required to demonstrate this link.


We thank the Biotechnology and Biological Sciences Research Council and Delta Biotechnology Ltd (now Novozymes) for an Advancement and Support of Education Award (to A.J.S.), the European Community for a Marie Curie Fellowship (to C.A.B.), the Engineering and Physical Sciences Research Council and BP (British Petroleum) for a Dorothy Hodgkin Postgraduate Award (to J.L.), the Wellcome Trust (Edinburgh Protein Interaction Centre), the Wolfson Foundation, and the Royal Society (Olga Kennard Fellowship to C.A.B.) for their support of this work.


  • Metal Metabolism: Transport, Development and Neurodegeneration: A Biochemical Society Focused Meeting held at Imperial College London, U.K., 9–10 July 2008. Organized and Edited by David Allsop (Lancaster, U.K.) and Harry McArdle (Rowett Research Institute, Aberdeen, U.K.).

Abbreviations: ATCUN, N-terminal copper and nickel-binding; HSA, human serum albumin


View Abstract