Berkelium (Bk) has been predicted to be the only transplutonium element able to exhibit both +III and +IV oxidation states in solution, but evidence of a stable oxidized Bk chelate has so far remained elusive. Here we describe the stabilization of the heaviest 4+ ion of the periodic table, under mild aqueous conditions, using a siderophore derivative. The resulting Bk(IV) complex exhibits luminescence via sensitization through an intramolecular antenna effect. This neutral Bk(IV) coordination compound is not sequestered by the protein siderocalin—a mammalian metal transporter—in contrast to the negatively charged species obtained with neighbouring trivalent actinides americium, curium and californium (Cf). The corresponding Cf(III)–ligand–protein ternary adduct was characterized by X-ray diffraction analysis. Combined with theoretical predictions, these data add significant insight to the field of transplutonium chemistry, and may lead to innovative Bk separation and purification processes.
Berkelium (Bk) chemistry is largely unexplored compared with other transplutonium elements encountered in measurable amounts in nuclear chemistry, namely americium (Am), curium (Cm), and californium (Cf). Element 97, Bk is a peculiar case in the actinide (An) series because it is ‘light’ enough to be formed by multiple neutron-capture processes during nuclear detonations or in today's nuclear fission reactors, to such extent that it raises concerns for nuclear waste management1,2. However, its only isotope available in bulk quantities, 249Bk, has a relatively short half-life of 330 days, which hinders its use in chemical studies. In contrast, 243Am (t1/2 = 7,370 yr) and 248Cm (t1/2 = 340,000 yr) are long-lived isotopes that can be produced and purified on the multi-gram scale. Likewise, Cf (Z = 98), although heavier than Bk, is easier to synthesize and investigate thanks to the relatively stable isotopes 249Cf (t1/2 = 351 yr) and 252Cf (t1/2 = 2.65 yr). A consequence of its extreme rarity as a laboratory material and its highly radioactive nature, very few Bk coordination compounds have been characterized to date3,4 and little is known about the behaviour of Bk in environmentally and biologically relevant species5,6. Due to the special stability afforded by the half-filled 5f7 electronic configuration, Bk is the element among the actinides heavier than plutonium (Pu) that can most readily be oxidized from oxidation state +III to +IV in aqueous systems; its neighbouring elements, Am, Cm, Cf and einsteinium (Es) having only limited or inexistent stability as M(IV) ions7. Even though rutherfordium (Rf, Z = 104) is expected to exhibit coordination chemistry properties similar to those of zirconium (Zr) and hafnium (Hf) and would therefore also readily form a stable M(IV) ion in solution8, its isotopes are short-lived (with half-lives from μs to ∼1.3 hours) and Bk is, until now, the heaviest element of the periodic table available for bulk chemistry that can be studied as a tetravalent ion. It preferentially exists in aqueous solutions as Bk(III) but oxidation to Bk(IV) under very drastic conditions was reported shortly after the discovery of the element9,10. The standard oxidation potential for the couple Bk4+/Bk3+ is +1.60 V (in 1 M HClO4 versus NHE)11, which makes the formation of Bk(IV) species arduous, yet feasible, with a large excess of very strong oxidizers such as bromate, bismuth trioxide and lead dioxide, or by prolonged electrolytic oxidation. The redox properties of Bk are relatively close to those of cerium (Ce, E°Ce4+/Ce3+ = +1.7 V) and differ significantly from what is observed with the corresponding lanthanide (Ln) ion, terbium (Tb, E°Tb4+/Tb3+ = +3.3 V) (ref. 12). This ambivalence, if adequately unlocked, would confer to Bk a chemical behaviour readily different from that of adjacent ions, Cm and Cf, whose only stable oxidation state in solution is +III.
Classical attempts to extend the chemistry of Bk have yielded a decrease of the Bk4+/Bk3+ potential to +1.56 V in 1 M HNO3 and +1.37 V in 1 M H2SO4 (refs 3,13). An almost identical trend is known for the couple Ce4+/Ce3+, whose redox potential is +1.70 V, +1.61 V, +1.44 V and +1.28 V in 1 M HClO4, HNO3, H2SO4 and HCl, respectively14. The difference in stability between the Bk(IV) and Bk(III) species formed with these common anions is, however, not strong enough to make Bk(IV) readily accessible. In fact, once tremendous efforts are made to oxidize Bk(III) in these mineral acids, Bk(IV) complexes are naturally reduced within a few hours or even minutes15,16,17. More exotic approaches to stabilize Bk(IV) have used saturated pyrophosphate solutions17, acidic mixtures of bromate and heteropolytungstate K10P2W17O61 (ref. 18), or triphenylarsine oxide in pure acetonitrile19, all transient and difficult systems to implement in separation processes or for the sequestration of Bk in vivo. A more realistic strategy was proposed in the early 1970s using highly concentrated carbonate solutions13,16, with a conditional oxidation potential Bk4+/Bk3+ of +0.26 V in 2 M K2CO3 (pH 10). Aside from requiring a tremendous excess of carbonate ions, this medium's drawbacks include a narrow pH-range under which Bk(IV) species are stable, the presence of multiple species with unknown stoichiometry or stability20, the limited solubility of carbonate salts, and its incompatibility with in vivo conditions or most nuclear waste treatment operations in acidic media.
Instead of using large excesses of inorganic complexing ions, we selected a water-soluble small organic molecule that is highly selective toward An(IV) ions in order to lift the thermodynamic barrier imposed by the high redox potential of the couple Bk4+/Bk3+. The synthetic siderophore analogue 3,4,3-LI(1,2-HOPO) (1, Fig. 1), composed of 1-hydroxy-pyridine-2-one (1,2-HOPO) units linked to a central linear spermine scaffold, is gaining attention as a therapeutic decorporation agent for its bio-compatibility, its low toxicity21,22,23, and its strong efficiency at sequestering actinide and lanthanide ions, including UO22+, Ln3+, An3+, Th4+ and Pu4+ (refs 24,
Results and Discussion
Bk redox potentials are usually 50–100 mV lower than their Ce counterparts3, the redox potential for [Bk(IV)1]/[Bk(III)1]− is expected to be around −0.1 V (versus NHE), suggesting an easy and direct access to Bk(IV) in solution. This estimated value was confirmed by density functional theory (DFT) calculations on both gas-phase and solvated Bk(III) and Bk(IV) species, for which coordinates and frequencies of the calculated species are listed in Supplementary Tables 2–7; in those solvated species, a Bk4+/Bk3+ electrochemical potential with an upper bound of −0.13 V relative to NHE was determined. Considering trends in actinide ionic radii29 and corresponding Th(IV) and Pu(IV) complex stability constants, a formation constant of 10+44.7 (at 25 °C) is estimated for [Bk(IV)1], a drastic increase over [Bk(III)DTPA]2– (DTPA = diethylenetriaminepentaacetic acid, log(β110) = 22.8) and [Bk(III)DPA3]3− (DPA = dipicolinic acid, log(β130) = 23.1), currently the most stable known Bk coordination compounds4,30. Rigorous metal- or ligand-competition assays commonly used to determine thermodynamic constants for other 4+ metal complexes cannot be applied to Bk(IV) as they would necessitate a competing ligand that could stabilize Bk(IV) once released by 1 with a known affinity for Bk(IV). Such a system is currently unavailable, because 1 is the first organic ligand reported to stabilize Bk(IV) and no thermodynamic stability constant for a Bk(IV) aqueous species has ever been established.
Photophysical properties of chelated Bk(IV)
Stabilization of Bk(IV) by 1 was first indicated by UV–visible spectrophotometry. A striking difference is observed between the An(III) and An(IV) complexes, with respective absorbance bands centred at 318 and 305 nm, as shown for Am(III) and Pu(IV) (Fig. 1c), and the Bk species displaying a maximum absorbance at 305 nm for this sharp band characteristic of the 1,2-HOPO π-π* transitions. In addition, the Bk complex exhibits a ligand-to-metal charge transfer band centred at 400 nm, a feature also observed with Pu(IV) and Ce(IV) complexes26,28. Absorbance spectrum assignment to Bk(IV) was further corroborated by theoretical calculations (Fig. 1d). Luminescence properties have long proven instrumental for actinide complex characterization but most studies focused on Cm species, owing to the 5f7 electronic configuration and intrinsically intense 6D7/2 → 8S7/2 emission of Cm(III), which facilitates direct excitation of f–f transitions. Luminescence studies of Bk species are scarce and limited to the solid-state31,32, with the exception of a report on Bk(III) luminescence in 0.5 M DCl (ref. 33). Characterization of [Bk(IV)1] provided the first Bk(IV) luminescence spectral features in aqueous solution via intramolecular sensitization through the so-called antenna effect: ligand excitation at 320 nm resulted in sharp emission peaks at 590, 612, 659 and 702 nm, with the most intense centred at 612 nm (Fig. 1e). The corresponding excitation spectrum monitoring emission at 612 nm displayed a main transition centred at 305 nm (Supplementary Fig. 1) and characteristic of energy transfer from the ligand 1,2-HOPO units27. The observed structured four-peak emission is due to ligand field splitting of the emitting state, J = 7/2, as the spherical symmetry of the half-filled 5f7 configuration should only result in a small splitting (945 cm−1 between the lowest and highest Stark levels for the free ion versus 2,700 cm−1 for the chelated ion). Interestingly, the bathochromic shift of the emission maxima typically observed during complexation of lanthanide and actinide ions appears very pronounced here: from the highest to the lowest Stark levels of the excited state 6D7/2, shifts of ∼10 to 90 nm (equivalent to energy shifts of up to 2,100 cm−1 or roughly 12%) were observed in the complexation of Bk by 1. While large shifts (on the order of ∼50 nm, or ∼1,100 cm−1) have been noted in 5f systems before34, the relative decrease in inter-electronic repulsion in this Bk species is remarkable, especially when compared with the corresponding Am(III) (ref. 27) and Cm(III) (ref. 25) species that displayed respective relative shifts of the order of 1 and 2.5%.
These features bestow additional evidence of the stabilization of Bk(IV) by 1 since, based-on the reported energy levels of Bk(III), the ligand cannot sensitize Bk(III) (Fig. 1f). Correspondingly, excitation of the ligand in [Cf(III)1]− did not result in any luminescence features, as expected from the energy levels of Cf(III). The main emission of [Bk(IV)1] (612 nm, Φ = 1.2 × 10–5) is consistent with the 611 fluorescence band reported for solid Bk-doped CeF4 after direct excitation31 and is close to, but completely distinct from, that reported for the isoelectronic Cm(III) complex (610 nm, Φ = 0.45) (ref. 25). Despite the similar half-filled 5f7 configurations of both metal ions, the stronger spin-orbit coupling associated with Bk(IV) (ζ = 3,244 for the free ion35) gives rise to metal-centred electronic energy levels closer to each other than in the case of Cm(III) (ζ = 2,889)36. Hence the much lower ligand-to-metal energy transfer efficiency observed for the Bk(IV) species is most likely due to the induced presence of a large number of non-emissive energy levels between the ligand triplet excited state and the metal 6D7/2 emitting state (Fig. 1f). The Bk(IV) complex displayed a bi-exponential luminescence decay (Supplementary Fig. 2) with two distinct lifetimes (188 ± 19 µs and 6.7 ± 0.9 µs), both increasing slowly but linearly with D2O content in solution (Supplementary Table 10) and shorter than but in the same range as that of the Cm(III) complex (383 ± 38 μs), consistent with a slightly smaller gap between the 6D7/2 emitting and 8S7/2 accepting levels for Bk.
Oxidation state confirmation by mass spectrometry
The Bk oxidation state when bound to 1 was unambiguously assigned through liquid chromatography (LC) coupled with high-resolution mass spectrometry (MS). Analysis of 1:1 metal:ligand aqueous mixtures prepared under ambient conditions with 241Am(III), 248Cm(III) and 249Cf(III), whose M4+/M3+ redox potentials are extremely high (+2.6, +3.1 and + 3.2 V, respectively)7,37 confirmed the formation of trivalent complexes of 1. For those three transplutonium elements, MS patterns are almost identical, with four mono-charged adducts detected ([M(III)LH2]+, [M(III)LHNa]+, [M(III)LNa2]+ and [M(III)LNaK]+), which clearly contrasts with the data obtained for tetravalent 242Pu and 232Th complexes (Fig. 2; Supplementary Fig. 3). The MS spectrum of the 249Bk system assembled in situ from a Bk(III)Cl3 solution displayed [BkLH]+, [BkLNa]+ and [BkLK]+ species, demonstrating that the resulting complex contains a Bk(IV) ion and not Bk(III). Spontaneous stabilization of Bk(IV) is thought to occur through air oxidation, similarly to the Ce system, which does not necessitate the addition of oxidizers or the electrolytic oxidation required in previously proposed methods. The use of 1 as a chelation and oxidation-promoting agent for Bk(III) also has the notable advantage of promoting the formation of M(IV) species over a wide pH-range: the Zr(IV), Ce(IV) and Pu(IV) complexes are formed in 1 M H2SO4 (ref. 26) and are stable up to pH 11 (Supplementary Fig. 4). Finally, the very high formation constants of the [M(IV)1] complexes prevent competition with any ligand potentially present in the media, which gives more flexibility and robustness to the system.
Siderocalin-based macromolecular recognition
Stabilization of Bk(IV) by 1 was further confirmed using secondary macromolecular recognition. The human protein siderocalin (Scn), known to intercept ferric complexes of microbial siderophores as an immune response against bacterial invasion and recently highlighted as a potential player in actinide mammalian uptake, specifically recognizes negatively charged Ln(III), Am(III) and Cm(III) complexes of 1 (ref. 38) through tight electrostatic interactions, but not the neutral Th(IV) and Pu(IV) analogues. A fluorescence quenching assay revealed Scn does not recognize Bk–1, evidencing the complex neutrality and indirectly confirming the stabilization of Bk(IV) and not Bk(III) by 1. In contrast, [Cf(III)1]–-bound Scn with an equilibrium dissociation constant Kd of 50 ± 5 nM, which is in the same range as those reported for Am(III) and Cm(III) (29 and 22 nM, respectively)38. The structure of the [249Cf(III)1]–/Scn ternary complex was determined (Supplementary Table 11) using methods analogous to those derived for corresponding lanthanide (Sm) and actinide (243Am, 248Cm) metal/chelator/Scn adducts38; it is the first crystallographic report of a macromolecular Cf assembly. Those prior structures demonstrated that octadentate complexes bind the trilobal Scn ligand binding site, or ‘calyx’, much as native ferric siderophore complexes do, with 1,2-HOPO units fitting snuggly into sub-pockets within the calyx (Fig. 3a). The overall 249Cf(III) structure is very similar, particularly with regard to the protein (Fig. 3b). However, in detail, the 1,2-HOPO subunits of the chelator showed much more variability among the three views of the [249Cf(III)1]– complex (z = 3) than among the 18 near-identical views (six per crystallographic asymmetric unit, z = 6) of the previous three complex structures (Sm, 243Am, 248Cm; Fig. 3c). The octadentate coordination around the Sm, 243Am, and 248Cm centres that was best described as a ‘snub disphenoid’ was altered in the case of 249Cf, resulting in incomplete coordination of the 249Cf(III) centre by two of the 1,2-HOPO groups; probably due to different bond lengths between metals and 1,2-HOPO groups constrained to bind within the rigid calyx (Supplementary Table 11). The coordination differences between the Sm and 249Cf structures are noteworthy, as both metal ions are comparably sized39, and suggest that the Cf 5f orbitals may participate in increased orbital mixing.
Protein- and ligand-based separation of Bk
Current processes to separate Bk from Am, Cm, Cf, and fission products after production necessitate numerous steps and use strong oxidizers such as bromate to segregate Bk(IV) from the non-tetravalent ions3. Liquid–liquid extraction steps involving mixed mineral acid and organic solvents as well as multiple final separation steps using ion-exchange columns are almost always included in those processes3,40, even in the case of more recently reported rapid separation techniques41. The non-recognition of [Bk(IV)1] by Scn suggests innovative procedures to separate Bk from M(III) ions could consist of passing a solution of the irradiated mixture complexed with 1 through a Scn-containing medium, followed by discrimination using either size, mass, affinity, polarity or solubility differences. However, the separation of Bk(IV) from other M(IV) ions potentially present during Bk production, namely Zr, Ce, Th and Pu, also presents a challenge. The Ce–Bk pair is currently known as the most difficult due to the almost-identical redox properties of the two elements3, which has led to complicated solvent extraction or ion exchange techniques42,43,44,45,46. Figure 3d displays the relative retention of various M(IV) complexes of 1 on a classical C18 LC column. The retention time of the Bk complex falls between those of Zr(IV) and Ce(IV), trending with the ionic radii of the octa-coordinated metals (0.84, 0.93, and 0.97 pm for Zr(IV), Bk(IV) and Ce(IV), respectively)29. Without actual optimization, [Bk(IV)1] was easily discriminated from its Ce, Th and Pu analogues. Hence, a two-step separation process is sufficient for isolating Bk from all other lanthanide and actinide ions, with step 1 sequestering 3+ ions based on Scn selectivity toward [M(III)1]– complexes, and step 2 separating Bk from 4+ ions under classical chromatography. Besides stabilizing the heaviest +IV ion of the periodic table available for bulk chemistry, this procedure represents tremendous progress for Bk chemistry as it keeps Bk species in aqueous phase throughout the process, is operated at room temperature and does not introduce additional non-volatile elements or require any biphasic liquid–liquid extraction step, with limiting variables such as solvent loading capacity, pH robustness, extractant solubility, third-phase formation, etc. Recovery of Bk from the final eluted solutions would be achieved either through acidification and ashing or by precipitation of Bk-hydroxide species upon pH increase and subsequent filtration. Both methods (ashing or precipitation), commonly used in hydrometallurgy and would allow recovery of solid Bk. Finally, the protein/ligand-based separation strategy described here could be used to design cleaner and softer methods for metal-ion sorting based on bioengineered systems.
249Cf (351 yr half-life; 4.1 Ci g−1) represents a serious health risk owing to its α emission (6.194 MeV), its γ emission (0.388 MeV), and emission from its decay products such as 245Cm (8,500 yr half-life), an α emitter (5.623 MeV) that undergoes spontaneous fission. 249Bk (330 d half-life) is a β emitter that decays to 249Cf. 248Cm (3.49 × 105 yr half-life; 5.162 MeV) and 242Pu (3.74 × 105 yr half-life ; 4.985 MeV) are α emitters. All these elements where manipulated in laboratories designed for the safe handling of transuranics.
Chemicals were acquired commercially and used as received. Ligand 1 was prepared and characterized as previously described47. Stock solutions (4 mM) of 1 were prepared by dissolution of a weighed portion of ligand in DMSO and aliquots were removed prior to each experiment. Aliquots of acidified stocks of carrier-free 249Cf and 248Cm from the Lawrence Berkeley National Laboratory were used. A stock solution of 249Bk(III) in 0.1 M HCl was prepared from solid 249BkCl3 obtained from the Oak Ridge National Laboratory. All measurements were completed within six weeks of the original separation work and within two weeks of dissolution of the dry salt. All aqueous solutions were prepared using deionized water purified by a Millipore Milli-Q reverse osmosis cartridge system and the pH was adjusted as needed with concentrated HCl or KOH. For direct spectroscopic measurements, equimolar amounts of metal and chelator were used to constitute complex solutions (40 μM, pH 8.4) in 0.1 M CHES buffer. Recombinant human Scn was prepared as previously described48.
Liquid chromatography–mass spectrometry
Liquid chromatography–high resolution mass spectra (LC–HRMS) were acquired on a UPLC Waters Xevo system interfaced with a QTOF mass spectrometer (Waters Corporation, Milford, Massachusetts, USA) in Micromass Z-spray geometry. The previously described experimental setting26 used for LC–HRMS assays is detailed in Supplementary Information and was applied to samples containing an equal concentration of actinide and 1 in 0.1 M HEPES buffer (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) at pH 7.4 (for Cm, Cf and Bk) or in 0.5% formic acid at pH 2 (for Ce, Th and Pu). Concentrations were 10 μM for 243Am, 249Bk and 249Cf samples and 1 μM for Ce, 232Th, 242Pu and 248Cm.
Protein fluorescence quenching binding assay
The affinity of Scn for complexes of 1 was quantified by monitoring the intrinsic fluorescence of the protein upon complex-binding, as described previously38 and detailed in Supplementary Information.
UV–visible absorption spectra were recorded on an Ocean Optics USB 4000 absorption spectrometer, using quartz cells of 1.00 cm path length. Emission spectra were acquired on a HORIBA Jobin Yvon IBH FluoroLog-3 spectrofluorimeter, used in steady state mode. Luminescence lifetimes were determined on a HORIBA Jobin Yvon IBH FluoroLog-3 spectrofluorimeter, adapted for time-correlated single photon counting (TCSPC) and multichannel scaling (MCS) measurements. Further experiment and instrument details are provided in Supplementary Information.
For crystallization, 1 mM solutions of equimolar [Cf(III)1]– complex were mixed in a 2:1 molar ratio with Scn, which was then buffer-exchanged into 25 mM PIPES (pH = 7.0), 150 mM NaCl, 1 mM EDTA, and 0.01% w/w NaN3, and concentrated to ∼10 mg ml–1 protein. Diffraction-quality crystals were grown by vapour diffusion as detailed in Supplementary Information, along with data collection and refinement methods. Crystallographic statistics are reported in Supplementary Table 11.
Computational chemistry simulations
All calculations were performed with the latest development version of the open-source NWChem software suite50. Scalar relativistic density functional theory calculations were carried out with the B3LYP density functional51,52, using the Stuttgart small-core effective core-potential and associated basis set for the actinide atoms53 and all-electron DFT optimized valence double-ζ polarized (DZVP) basis sets54 for the light atoms in the complex. UV–visible spectra were calculated at the time-dependent density functional theory (TDDFT) level of theory55. TDDFT equations were solved using a novel new symmetric Lanczos algorithm56. Geometries of solutions species were optimized within the COSMO model. The DIRAC code57 and Dyall's relativistic triple-zeta basis set58 were used to estimate the effect of atomic spin-orbit coupling on Bk(IV) with a 5f 7 8S7/2 ground state, and Bk(III) with a 5f 8 7F6 ground state at the Dirac–Hartree–Fock level of theory. For each atom an average-of-configurations SCF was performed followed by a Complete Active Space CI (COSCI) in the space covered by the 5f 7 or 5f 8 to project out all the states. Additional computational details are provided as Supplementary Information.
The data supporting the findings discussed here are available within the paper and its Supplementary Information files, or from the corresponding authors upon request. The final models for the described protein structure have been deposited in the PDB49, under accession code 5KIC.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Protein Data Bank
This work was supported by the US Department of Energy (DoE), Office of Science Early Career Research Program and Office of Science, Office of Basic Energy Sciences, Chemical Sciences, Geosciences, and Biosciences Division at the Lawrence Berkeley National Laboratory under contract DE-AC02-05CH11231 (R.J.A.), by the National Institutes of Health (R01DK073462, R.K.S.), and by the Scientific Discovery through Advanced Computing (SciDAC) program of the US DoE, Office of Science, Office of Advanced Scientific Computing and Office of Basic Energy Sciences (W.A.d.J.). The Radiochemical Engineering and Development Center at Oak Ridge National Laboratory is supported by the US DoE, Isotope Development and Production for Research and Applications Program. The Advanced Light Source (ALS) and Energy Research Scientific Computing Center (NERSC) are supported by the Director, Office of Science, and Office of Basic Energy Sciences, of the US DoE under contract no. DE-AC02-05CH11231. The Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program provided an award of computer time through the Oak Ridge Leadership Computing Facility, a US DoE Office of Science User Facility supported under Contract DE-AC05-00OR22725. We thank M. Allaire, S. Morton, J. Bramble, K. Engle, M. Dupray, and I. Tadesse for assistance in implementing diffraction data collection on radioactive crystals at ALS 5.0.2 beamline.
About this article
Nature Chemistry (2017)