Heavy Atom Compound Choices

I've restricted this list to the commoner heavy atoms, adding exotics as information becomes available. 1 Å is approx 12.3985 keV. CuKα (1.5418 Å) is approx 8.041 keV.

A general description of the heavy atom screening process can be found in Boggon and Shapiro (2000) Structure 8, R143-R149 and expansion on some overlooked aspects of characterization can be found in Garman and Murray (2003) Acta Cryst D59, 1903-1913 at http://journals.iucr.org/d/issues/2003/11/00/ba5042/index.html. There's also a good table for anomalous scatterers at CuKα and 1.08 Å from Scripps.

Covalent Modifiers

These compounds - almost invariably transition metals - bind to His, Cys and Met groups within the protein with semi-covalent interactions. Third row transition metals are usually preferred because they are electron-dense but sometimes second row transition metals have been used. Affinities of the bare ions for the target groups (e.g. Hg2+ in HgCl2) can result in partial protein denaturation as the ion burrows into the protein and disrupts the fold. In consequence the affinities are often "turned down" by using compounds that are already covalently linked to one or more groups. With some transition metal ions, use of ligands with higher affinity than water (bare ions assumed to be something like M(H2O)6 in solution), like NH3, CN-, SCN- can also reduce reactivity by being harder to displace than H2O. In the case of Platinum, the Pt-IV oxidation state is less reactive than the Pt-II oxidation state. Mercury is often linked to bulky organic groups (e.g. in PCMBS, PHMBS, EthylHgCl) which limits its ability to penetrate a protein molecule and attach to internal Cys and disrupt the fold. Trimethyl Lead is a non-transition metal that shows some covalent-type affinities and behaves quite differently from the bare Pb2+ ion.

No corrections are made for oxidation state in the f'' tables below and in general the values will be shifted from these - I extracted these data from Ethan Merrit's site at http://skuld.bmsc.washington.edu/scatter/AS_periodic.html.

Compound Group Targets Edge f'' at 1.0 Å and CuKα
HgCl2 Hg2+ Cys, His L-III, 12.284 keV, 1.009 Å, f'' 10, broad 10 and 7.7
AuCl3 Au3+ His, Cys L-III, 11.919 keV, 1.040 Å, f'' 10 9.1 and 6.9
PtCl4 Pt2+ His, Met, Cys L-III, 11.564 keV, 1.072 Å, f'' 10 9.1 and 6.9
PtCl6 Pt4+/Pt2+ His, Met, Cys L-III, 11.564 keV, 1.072 Å, f'' 10 9.1 and 6.9
PbMethyl3 Pb2+ His, Cys L-III, 13.035 keV, 0.951 Å, f'' 10.1 4.3 and 8.5

Substitutions and Naturally Occurring Ions

Classically Seleno-methionine can substitute for S-Methionine with relatively little purturbation of the structure and in some cases mutation of Leu to Met can provide additional sites. Oxidation of SeMet is sometimes a problem, usually kept in check with DTT or β-mercaptoethanol, but alternatively one can force oxidation of SeMet - see Sharff et al. (2000) Acta Cryst D56, 785-788 at http://journals.iucr.org/d/issues/2000/06/00/li0349/index.html. Because SeMet is not an epic anomalous scatterer you still need 1 per 40-50 amino acids (more, with weakly diffracting crystals) - although Boggon and Shapiro suggest 1 per 100 is viable with care during data collection. However I have successfully phased off the Zn2+ anomalous signal which is fairly comparable, at a relatively impressive 1 Zn per 450 amino acids. Of course that Zn was tightly bound and 100% occupied with a crystal that diffracted fairly well. First row transition metals are moderately frequent in protein structures and those in the range Fe-Zn have accessible L-III edges at most synchrotron beamlines. They also have small anomalous signals at 1.0 Å. Mn is a bit of a problem, however as its edge is at nearly 1.9 Å. Rarely, Cacodylate can derivatize Cys residues to form Cys-S-AsO(Methyl)2 so I have included the data for Arsenic as well - a few groups have used this as a phasing method.

Compound Group Targets Edge f'' at 1.0 Å and CuKα
SeMet Se (Met sites) K, 12.658 keV, 0.980 Å, f'' 3.8 1.1 and 0.5
Zinc Zn2+ (Metal sites His/Cys) K, 9.659 keV, 1.284 Å, f'' 3.9 2.6 and 0.7
Manganese Mn2+ (Metal sites Asp/Glu/His) K, 6.539 keV, 1.896 Å, f'' 4.0 2.7 and 1.4
Arsenic As3+ (Cys via cacodylate) K, 11.867 keV, 1.045 Å, f'' 3.9 3.5 and 1.0

If your metalloprotein crystal diffracts really well you can phase the structure off metal ions with inaccessible edges simply by exploiting the significant f'' signal at the high energy end of the absorption edge. For the structure of Cadmium-bound marine diatom carbonic anhydrase (see Xu et al 2008) I used the Cd f'' signal at 1.6 Å well above the L-I 3.1 Å absorption edge to successfully SAD phase the structure with pretty good phases with about 4 electrons of anomalous scattering. The fact that the crystal diffracted to better than 1.5 Å obviously helped here. Although not really a heavy atom derivative there has been real progress with Sulfur-SAD structures (see for example this article on Sulfur-SAD with Cr Kα radiation).

Non-covalent Modifiers

Under some circumstances the cyanide derivatives of Hg, Pt, Au act essentially as anions and fall into this class. Cyanide is often extremely difficult to displace from a transition metal ion. However we will assume that the table below is mainly plain cations that bind to proteins via electrostatic (and possibly h-bond) interactions. Over the years this has tended to be UO22+, Pb2+ and the Lanthanides. The Lanthanides are getting closer scrutiny because they have rather considerable f'' values at their L-III edges and still pretty substantial f'' values at 1.0 Å. A few of them are pretty good at CuKα too.

Zbigniew Dauter et al. has pioneered the use of Bromide (Br-) and Iodide (I-) for potential quick-and-dirty derivitization for SAD/MAD data collection at synchrotron beamlines (Br) or MIR/SIRAS at home sources (I). In my hands this hasn't really paid off and you need existing phases of at least mediocre quality to even find the Br or I atoms but there are at least some structures which have been phased this way.

Lanthanide complexes have been getting rather more attention lately. Lanthanides replace Calcium in certain circumstances - I think William Weiss's Holmium derivative of Mannose Binding Protein might be the first example of that in a MAD context, back in the early 1990's, but recent papers suggest compounds like Gd-HPDO3A might be more general heavy atom derivatives with strong anomalous scattering properties. The basic problem with some of the Lanthanides is that their L-III edge is up beyond CuKα, often too low energy to be accessible for many synchrotron beamlines, but Ytterbium is better in this regard. See Girard et al (2002) Acta Cryst D58, 1-9 at http://journals.iucr.org/d/issues/2002/01/00/li0415/index.html. Some of the Lanthanides have a pretty good anomalous signal at CuKα if your crystal diffracts reasonably well (i.e. Sm).

Xenon and Krypton have also been used as derivatives despite the fact that they are monatomic noble gasses and therefore do not undergo any chemical reactions - you can force them in your crystal in some cases by the use of pressurization. Xe is the heavier of the two, with good f'' at CuKΑ, but Kr has an accessible MAD edge. They tend to bind in the hydrophobic core of proteins.

Uranium compounds, if you can still buy it without attracting the attention of the FBI, are still good heavy atoms, with good f'' values at CuKα and 1.0 Å. U6+ is so polarising it strips the protons off waters and the relevant ion is the UO22+ group. It also has an epic MIR phasing signal, if you can find an isomorphous native.
Compound Group Targets Edge f'' at 1.0 Å and CuKα
UO2(NO3)2 UO22+ Glu/Asp etc inaccessible 6.9 and 13.4
Pb(OAc)2 Pb2+ Glu/Asp etc L-III, 13.035 keV, 0.951 Å, f'' 10.1 4.3 and 8.5
Sm(OAc)3 Sm3+ Glu/Asp, Ca sites L-III, 6.716 keV, 1.846 Å, f'' 10.6 6.1 and 12.1
Ho(OAc)3 Ho3+ Glu/Asp, Ca sites L-III, 8.071 keV, 1.536 Å, f'' 10.6 8.2 and 3.7
Yb(OAc)3 Yb3+ Glu/Asp, Ca sites L-III, 8.944 keV, 1.386 Å, f'' 10.5 9.7 and 4.4
NaBr Br- Arg,Lys etc K, 13.474 keV, 0.920 Å, f'' 3.8 0.6 and 1.3
NaI I- Arg,Lys etc inaccessible 3.3 and 6.8
Kr Kr none K, 14.326 keV, 0.866 Å, 3.8 0.7 and 1.4
Xe Xe none inaccessible 3.6 and 7.3


Clusters have the advantage that they magnify the scattering due to a single heavy atom binding to a site. Ta6Br12 has been fairly widely used as a derivative. The tremendous advantage is the enormous f'' signal at low resolution, however this signal drops precipitously to near zero around 6Å and is rather more modest at higher resolutions. The cluster is also often rotationally disordered with the crystal since it binds with less-specific electrostatic interactions. Nevertheless it was quite invaluable as part of the structure solution of the PP2A ABC holoenzyme structure, where we lacked a good model for the B subunit and the low resolution Ta6Br12 phases were enough to place the WD40 domain with enough precision to start building. (See: A selection of reference on the Ta6Br12 cluster are:

Others, particularly people working with the huge ribosome structures, have exploited other heavy atom clusters. For example see: Thygesen J, Weinstein S, Franceschi F, Yonath A. "The suitability of multi-metal clusters for phasing in crystallography of large macromolecular assemblies." (1996) Structure 4: 513-8.

Most recently some novel compounds have been introduced, including the I3C "magic triangle" developed by Tobias Beck: "A magic triangle for experimental phasing of macromolecules." Beck, T., Krasauskas, A., Gruene, T. & Sheldrick, G.M.(2008) Acta Crystallogr. Section D 2008, 64:1179-1182. Link to PDF of article via Tobias's site.

Heavy Atom Derivatives used for Membrane Protein Structures

See "Membrane's Eleven: heavy-atom derivatives of membrane-protein crystals." J. P. Morth, T. L. Sørensen and P. Nissen (2006) Acta Cryst D62, 877-862.

"Rational" Prediction of Heavy Atom Binding

Published as Sugahara et al (2005) Acta Cryst D61, 1302-1305 and available online. The associated website is: http://www.rsgi.riken.go.jp/.

Alternatively you can do your own leg-work at browse the Heavy Atom Databank at http://www.sbg.bio.ic.ac.uk/had/heavyatom.html.

Assorted References Related to this Subject