Heavy atom derivatives

This page will be used to accumulate information about heavy atom derivatives. Its purpose is to understand the binding behaviour of different heavy atom compounds to proteins & DNA. The information in these pages is mostly derived from:
-Chapter 8 of Blundell & Johnson
-Chapter 13 of Methods in Enzymology Vol. 114 by Gregory A. Petsko
-A handout of Prof. Jan Drenth

Another site with information on heavy atom use in protein crystallography plus further links is the Heavy Atom Databank

Heavy atoms are very toxic, so prepare yourself before you prepare derivatives by talking to experienced colleagues and/or reading what some real experts have to say.

Many texts divide the heavy atom reagents in class A and class B metals. These metals differ in their preference for hard (carboxylates and other oxygen containing groups) and soft (sulphur, nitrogen and halides) ligands respectively. Here I will divide potential heavy atom reagents in 6 groups based on their prefered protein interactions. This will cause some metals to appear in two or more groups but I think this makes more sense since some metals can behave very differently depending on their ligands.

Class A
Prefers O and F ligands. Low pH and AS (ammonium sulphate) no problem. Solubility problems at higher pH and with phosphate
Class B
Prefers N, S and halide ligands. Phosphate and higher pH less problem. Reduced pH (<6) and AS at higher pH (>7) unfavourable
"Inert" metal complexes with an overall negative charge. Binds to basic regions on a protein
"Inert" metal complexes with an overall positive charge. Binds to acidic regions on a protein
"Inert" metal complexes with a hydrophobic nature Binds to hydrophobic pockets
Iodination, selenomethionine

Class A metals

The class A metals typically consist of the alkali metals (Li+, Na+, K+, Rb+, Cs+), the alkaline earth metals (Mg2+, Ca2+, Sr2+, Ba2+), the
lanthanides (La3+, Ce3+, Nd3+, Sm3+, Eu3+, Gd3+, Dy3+, Er3+, Lu3+) and the actinide UO2(2+). In addition to these, several metals have a mixed class A and class B character and can be coordinated in a similar manner as the true class A metals (Tl+, Pb2+ and perhaps also cadmium).

The class A metals are not very polarizable and they bind hard, electronegative, ligands like F-, OH-, H2O, phosphate and carboxylates. The binding is best characterized as "electrostatic". Accordingly, in mother liquors with high ionic strength the interactions with class A metals is weakened. Class A metals are also unpractical at higher pH (pH >7.5) since the insoluble hydroxides are formed. A similar problem exists with phosphate and citrate which compete for the class A metals.

Due to the low pKa of the prefered ligands, class A metals can be used at low pH (approx. pH=4.0, but lower if there are Asp or Glu residues with depressed pKa values). They also do not suffer from competition with NH3 like many class B metals. In addition several heavy class A metals can substitute for their lighter counterparts, NA+/K+ or Mg2+/Ca2+.


Uranium is the heaviest metal in our repetoir and it is normally used as the linear (O=U=O)2+ uranyl ion. UO2Ac2, UO2(NO3)2 and K3UO2F5 are the most commonly used derivatives. The acetate derivative has the highest solubility but even the more reactive nitrate derivative can reach 100mM concentrations. The fluoride is least reactive and can be tried if the others give too much disorder or too many sites. A problem with uranyl derivatives appears to be that they are less specific than the lanthanides and often give clusters of sites with low occupancy. Please also keep in mind, that uranium is not only toxic but also radioactive.


The lanthanides form a series of metals for which the most stable oxidation state is III (eg. a +3 valency). Their ionic radius decreases from La3+ to Lu3+ and this determines to a large extend their specificity. The lanthanides are often used to replace Mg2+ or Ca2+ ions. It is often favourable to try several lanthanides. For instance, when replacing Ca2+ in thermolysin it was found that Er3+ and Lu3+, both of which have an ionic radius that is a bit smaller than Ca2+, gave the most isomorphous crystals. Sm3+ (samarium) is an attractive ion since it gives a large anomalous signal.

Thallium & Lead

Thallium and lead have a mixed class A and class B character. However, they are most class A-like. Tl+ can sometimes replace Na+/K+ and it may be attractive for this capability. It can also be used at high pH unlike many other class A metals (a pH of 9.1 has been reported). It should however be mentioned that thallium is VERY toxic.
Lead can bind to carboxylates, but it has also been reported to bind to imidazole when it substituted for zinc in insulin. Based on information sent to me by Gilles Precigoux it seems that cadmium also has a mixed A/B character.

Class B metals

The class B metals are all polarizable and prefer soft polarizable groups as ligands. Several of the most common derivatives belong to this class (eg. Pt2+ and Hg2+). These derivatives do not work well at low pH since their potential ligands, cysteines and histidines, will be protonated and less reactive. Pt is an exception since it can still interact with methionine and disulphide bridges at low pH. At higher pH (>7) and with AS as a precipitant they also do not function well since NH3 will compete for the metal. Phosphate and hydroxide are hard ligands and therefore do not bind tightly. Accordingly, phosphate buffers or elevated pH are less problematic than with the class A metals. Also, since the metal-protein interaction is (partially) covalent in nature, the binding is less sensitive to high ionic strength. However, for the halide salt, the reactivity of the compound can be reduced by increasing the concentration of the halide.
Under conditions where class B metals do not function well, one can still prepare stable complexes which bind through the properties of the complex;
anionic, cationic or hydrophobic.

Hg2+, mercury

There is an almost infinite list of mercury compounds that can be used as derivatives. They all share the binding preference for cysteine and histidine in their deprotonated states. The binding is often very tight and covalent in nature. In addition to the "general" mercury compounds, it is often possible to mercurate carbon compounds that bind specifically to your protein (eg. carbohydrates, inhibitors, substrates).
Many structures have been derivatized with "simple' mercury salts like: HgCl2, HgBr2, HgAc2. These are very small and can therefore reach less accessible surface sites. The chloride salt is less stable and therefore more reactive than the corresponding bromide (acetate is even more reactive??? because it is a hard ligand). These reagents can be made less reactive by increasing the concentration of the anion. Hg2+ added as a mercury salt can also replace some biometals like zinc, copper and perhaps some other.
The binding of mercury can be made more restrictive by complexing the metal to a larger organic compound. This limits the number of accessible sites. Some complexes also contain multiple mercury atoms to increase the scattering mass. In addition, the organic compound may contribute to the binding affinity. In the extreme case, the organic compound is chosen to have known specific interactions with the protein of choice. In this situation mercury may not interact with the protein at all. For a list of compounds see Table 8.VI in Blundell and Johnson.
If the small mercury salts do not appear to bind then all hope is not yet lost. Mercury derivatives with
hydrophobic, cationic character can be prepared to probe other sites on the protein. In addition, it may be possible to mercurate disulphide bonds. The disulphide bond will have to be reduced before it becomes reactive. This can be done with DTT or other reducing agents (be aware that sulphur containing reducing agents will bind tightly to mercury as well). Alternatively, it has been reported that (Hg2)2+ can both reduce and mercurate the disulphide bridge (Blundell & Johnson, p218).

Ag+, silver

Silver behaves somewhat similar to mercury but it appears to have a larger preference for histidine. In addition it is less reactive than mercury. Accordingly, if a proton has to be displaced in the reaction than mercury will do so at lower pH than silver. However, if mercury reacts too vigorously one can try silver. The most commonly used compound is AgNO3. Remember that Ag+ is very insoluble with Cl- and other halides.

Pt2+, platinum

Platinum is one of the top-performers with respect to number of derivatized crystals. K2PtCl4 is by many considered to be the most successful reagent. The Pt2+ complexes have a square planar geometry. There also exists a Pt4+ oxidation state which has an octahedral coordination, but these compounds tend to reduce to the divalent oxidation state under normal conditions. It should be noted that many Pt compounds are light sensitive and soaking experiments should be stored in the dark.
In addition to cysteine and histidine, Pt2+ also binds to the methionine sulphur and the sulphurs in a disulphide bridge (without need for reduction). Although Pt2+, like the other class B metals, is less effective at low pH this does not apply to methionine and disulphide binding. Therefore Pt2+ used at high and low pH may give rise to different derivatization. It has also been noted that PT2+ reacts faster with methionine and cysteine than histidine. Accordingly, one can obtain different derivatization by controlling the soaking time. One should also be aware that in presence of either phosphate or NH3, (PtCl4)2- can rearrange to form PtCl3(PO4)4-, Pt(NH3)2Cl2, Pt(NH3)Cl2(PO4)3- or (Pt(NH3)4)2+ (see Blundell & Johnson, p228). With longer soaking times these newly formed compounds may lead to additional binding sites. The promiscuity and flexibility of Pt2+ compounds makes it likely that one can find some condition in which binding is obtained. However, greater care must be taken to be able to reproduce the soaking experiment.
The most common platinum reagent is K2PtCl4 or the less reactive heavier halides or nitrite (K2Pt(NO2)4). Tetra-valent platinum in the form of K2PtCl6 or its bromide and iodide variants have been used as well. The stable (Pt(CN)4)2- and (Pt(NH3)4)2+ compounds can be used as species that bind through their anionic or cationic charge rather than through direct binding by the metal. The Pt(NH3)2Cl2 compound is neutral and can penetrate proteins to some extend.
If the platinum compound is too reactive one can try to use a more stable variant (eg used brominated or iodinated instead of chlorinated compounds). One can also switch to a gold compound, see below. Other options are to reduce the pH, concentration, soaking time or temperature or increase the concentration of the anion.

Au3+, gold

Gold behaves like platinum but it is less reactive and less stable. One can also use monovalent gold, KAu(CN)2, as a stable linear anionic species.

Pd2+, palladium

Palladium coordination chemistry resembles platinum and gold, however, this metal is more reactive and can be tried if the other metals do not react or do so only very slowly. K2PdCl4 and its heavier halide counterparts have been used.

Ir3+, Iridium

Iridium behaves like platium, including its sensitivity to light. However, iridium prefers an octahedral coordination. K3IrCl6 can bind to histidines by nucleophilic substitution of a cloride by the imidazole. The (IrCl6)3- ion can however also bind as an anion to basic protein regions. In contrast, (Ir(NH3)6)3+ may bind to acidic regions on the protein.

Os4+, Osmium

Osmium also behaves like iridium and it likewise forms octahedral complexes. The (OsCl6)2- compound is relatively stable and can bind as an anion to basic regions on the protein. In its octa-valent oxidation state K2OsO4 can read with cis-diols. This has been exploited to label t-RNA. Please be aware that this is a very aggresive reagent and epithelial cells, especially eyes, are very vulnerable.

Cd2+, Cadmium

I am not sure at this moment where cadmium should go. But Blundell & Johnson describe that cadmium, as well as mercury, can substitute for zinc in insulin. Since zinc binding involves histidine I have placed cadmium under the class B metals. However, it may have, like zinc, some class A character. If you know, please tell me.

Gilles Precigoux sent me the following insights on cadmium binding to ferritins:

[edited] Cadmium sulfate was used to crystallize ferritins and it preferentially
binds to Asp & Glu, but binding to His and Cys was also observed. Strong binding
sites are those involving more than one carboxylate residue. Binding sites with
only one ligand seems to be possible but the site occupancy factor is much
lower: 0.2-0.3. In the ferritin crystallization, cadmium was required as it
makes intermolecular salt-bridges. {I believe this has been seen more often
for cadmium as well as zinc; Bart}. In this study the pH of the solution was

Anionic heavy atom derivatives

In this category I wish to include all heavy atoms or derivatives thereof that predominantly interact with the protein through their overall negative charge. Accordingly, these compounds will be less effective at high ionic strength and in the presence of other anions which may compete for the same site. In addition, binding will be enhanced at lower pH where the protein will be more positively charged. Especially protonation of histidines may play a role. The compounds in this group comprise iodide and stable anionic metal complexes. The strong cyanide ligand will never be substituted from class B metal by protein-derived groups and therefore will form stable anionic complexes. Depending on the motherliquor and protein, weaker ligands like I-, Br- or even Cl- may also yield stable complexes. Common compounds in this class are: I-, (HgI3)-, (Pt(CN)4)2-, (IrCl6)3-, (Au(CN)2)-. It should be remembered that the highly polarizable iodide ion has a hydrophobic character in addition to its charge. I- and (HgI3)- can therefore also bind to hydrophobic pockets, see below.

Cationic heavy atom derivatives

In this category I wish to include all heavy atoms or derivatives thereof that predominantly interact with the protein through their overall positive charge. Accordingly, these compounds will be less effective at high ionic strength and in the presence of other cations which may compete for the same site. In addition, binding will be enhanced at higher pH where the protein will be more negatively charged. Especially deprotonation of histidines may play a role. The compounds in this group comprise the stable cationic metal complexes. The most common ligand in this class is NH3 and potential derivatives are: (Pt(NH3)4)2+, (Ir(NH3)6)3+, (Hg(NH3)2)2+ ((Au(NH3)4)3+ ???)

Hydrophobic heavy atom derivatives

Most derivatives are limited to binding sites on the protein surface due to their charged nature or their large size. However, several compounds have been reported to be able to penetrate the hydrophobic protein interior and could therefore be interesting compounds to test. The only truely hydrophobic compound is the noble gas Xenon which can occupy hydrophobic pockets. Xenon has to be added under pressure (2 Atm or more) which presents some technical problems. A special device has been described by Schiltz et al. (J. Appl. Cryst. 27, 950-960, 1994). An alternative procedure has been to produce liquid Xenon by immersing the X-ray capillary in liquid N2. After sealing the capillary the evaporating liquid Xenon reaches a sufficient pressure to derivatize the crystal (or explode the capillary :-). For references see Schiltz et al. A commercial device is now being sold by MSC to make nobel gas derivatives for cryocrystallography.

Other compounds with a hydrophobic nature are:

Ethyl mercury chloride or ethyl mercury phosphate
the latter is the favourite of Prof. Petsko but cannot be used in phosphate buffers.
This compound is formed in approx. 24 hours from (PtCl4)2- in AS precipitants at above neutral pH.
This compound is formed from K2HgI4. Its formation can be stimulated by addition of excess KI.
This anion has sufficient hydrophobic character to enter proteins.

Other heavy atom "derivatives"

This is the obligatory repository for the elements that do not fit any of the above groups. At the moment there are two entries. Iodide can be used in a chemical reaction to mono- or di-iodinate tyrosine residues. For this reaction approx. 0.4M KI+I2 is used. Read the literature about the pros and cons of this technique. Just note that iodination in presence of ammonium can give an explosive cocktail.
The other outlier is selenium which is now routinely incorporated into proteins as selenomethionine for MAD phasing.

Version: February 7, 1997
By: Bart Hazes

University of Alberta
Department of Medical Microbiology & Immunology
1-15 Medical Sciences Buiding
Edmonton, Alberta, Canada, T6G 2H7

For questions or suggestions please send e-mail to: bart.hazes@ualberta.ca