Enhanced rare

Nov 02, 2023

Nature volume 618, pages 87–93 (2023)Cite this article

2865 Accesses

127 Altmetric

Metrics details

Technologically critical rare-earth elements are notoriously difficult to separate, owing to their subtle differences in ionic radius and coordination number1,2,3. The natural lanthanide-binding protein lanmodulin (LanM)4,5 is a sustainable alternative to conventional solvent-extraction-based separation6. Here we characterize a new LanM, from Hansschlegelia quercus (Hans-LanM), with an oligomeric state sensitive to rare-earth ionic radius, the lanthanum(III)-induced dimer being >100-fold tighter than the dysprosium(III)-induced dimer. X-ray crystal structures illustrate how picometre-scale differences in radius between lanthanum(III) and dysprosium(III) are propagated to Hans-LanM's quaternary structure through a carboxylate shift that rearranges a second-sphere hydrogen-bonding network. Comparison to the prototypal LanM from Methylorubrum extorquens reveals distinct metal coordination strategies, rationalizing Hans-LanM's greater selectivity within the rare-earth elements. Finally, structure-guided mutagenesis of a key residue at the Hans-LanM dimer interface modulates dimerization in solution and enables single-stage, column-based separation of a neodymium(III)/dysprosium(III) mixture to >98% individual element purities. This work showcases the natural diversity of selective lanthanide recognition motifs, and it reveals rare-earth-sensitive dimerization as a biological principle by which to tune the performance of biomolecule-based separation processes.

The irreplaceable roles of rare-earth (RE) elements in ubiquitous modern technologies ranging from permanent magnets to light-emitting diodes and phosphors have renewed interest in one of the grand challenges of separation science—efficient separation of lanthanides1. The separation of these 15 elements is complicated by the similar physicochemical properties of their predominating +III ions, with ionic radii decreasing only 0.19 Å between LaIII and LuIII (ref. 7), which also leads to these metals co-occurring in RE-bearing minerals. Conventional hydrometallurgical liquid–liquid extraction methods for RE production utilize organic solvents such as kerosene and toxic phosphonate extractants and require dozens or even hundreds of stages to achieve high-purity individual RE oxides3,8. The inefficiency and large environmental impact of RE separations9 have stimulated research efforts into alternative ligands with larger separation factors between adjacent REs10,11,12,13,14, and greener process designs to achieve RE separation in fewer stages15 and using all-aqueous chemistry6,16,17,18,19,20.

The discovery of the founding member of the LanM family of lanthanide-binding proteins demonstrated that nature has evolved macromolecules surpassing the selectivity of synthetic f-element chelators4. The prototypal LanM, from M. extorquens AM1 (Mex-LanM), is a small (12-kDa), monomeric protein that undergoes a selective conformational response to picomolar concentrations of lanthanides4,18 and actinides21,22,23,24, has facilitated understanding of lanthanide uptake in methylotrophs25, and has served as a technology platform for f-element detection26, recovery18,27 and separation6. Unusually among RE chelators, Mex-LanM favours the larger and more abundant light REs (LREs), especially LaIII–SmIII, over heavy REs (HREs)4. Our recent demonstration that even single substitutions to the metal-binding motifs of Mex-LanM can improve actinide/lanthanide separations23 spurred us to investigate whether orthologues of Mex-LanM might possess distinct, and potentially useful, metal selectivity trends.

Herein, we report that the LanM from Hansschlegelia quercus (Hans-LanM), a methylotrophic bacterium isolated from English oak buds28, exhibits enhanced RE separation capacity relative to Mex-LanM. Whereas Mex-LanM is always monomeric, Hans-LanM exists in a monomer/dimer equilibrium, the position of which depends on the specific RE bound. Three X-ray crystal structures of LanMs and structure-guided mutagenesis explain Hans-LanM's RE-dependent oligomeric state and its greater separation capacity than that of Mex-LanM. Finally, we leverage these findings to achieve single-stage Hans-LanM-based separation of the critical neodymium/dysprosium pair. These results illustrate how intermolecular interactions—common in proteins but rare in small molecules—may be exploited to improve RE separations.

We have proposed4 several hallmarks of a LanM. First, LanMs possess four EF-hand motifs. EF hands comprise 12-residue, carboxylate-rich metal-binding loops flanked by α-helices, which traditionally respond to CaII binding;29 in Mex-LanM, however, EF hands 1–3 bind lanthanide(III) ions with low-picomolar affinity and 108-fold selectivity over CaII, resulting in a large, lanthanide-selective disorder-to-order conformational transition4. EF4 binds with only micromolar affinity. Second, adjacent EF hands in LanMs are separated by 12–13 residues—rather than the typical ≈25 residues in CaII-responsive EF-hand proteins—resulting in an unusual three-helix bundle architecture with the metal-binding sites on the periphery5. Third, at least one EF hand contains proline at the second position (in Mex-LanM, all four EF hands feature P2 residues). We searched sequence databases using the first two criteria and a sequence length of <200 residues, identifying 696 putative LanMs. These sequences were visualized using a sequence similarity network30 to identify LanM sequences that cluster separately from Mex-LanM. Notably, at a 65% identity threshold, a small cluster of sequences that is remote from the main cluster of 642 sequences is formed (Fig. 1a). This exclusive cluster (the Hans cluster), includes bacteria from several genera, including Hansschlegelia and Xanthobacter (Extended Data Fig. 1), all of which are facultative methylotrophs31.

a, Sequence similarity network of core LanM sequences indicates that Hans-LanM forms a distinct cluster. The sequence similarity network includes 696 LanM sequences connected with 48,647 edges, thresholded at a BLAST E value of 1 × 10−5 and 65% sequence identity. The black box encloses nodes clustered with Hans-LanM. The LanM sequence associated with Mex (downtriangle) and four within Hansschlegelia (uptriangle) are enlarged compared to other nodes (circles). Colours of the nodes represent the family from which the sequences originate. b, Comparison of the sequences of the four EF hands of Mex- and Hans-LanMs. Residues canonically involved in metal binding in EF hands are in blue; Pro residues are in purple. c, Circular dichroism spectra from a representative titration of Hans-LanM with LaIII, showing the metal-associated conformational response increasing helicity; apoprotein is bold black, LaIII-saturated protein is bold red. d, Circular dichroism titration of Hans-LanM with LaIII, NdIII and DyIII (pH 5.0). Each point represents the mean ± s.d. from three independent experiments. e, Comparison of Kd,app values (pH 5.0) for Mex-LanM (black18) and Hans-LanM (red), plotted versus ionic radius7. Mean ± s.e.m. from three independent experiments.

Source Data

Hans-LanM features low (33%) sequence identity with Mex-LanM (Supplementary Fig. 1) and divergent EF-hand motifs, particularly at the first, second and ninth positions (Fig. 1b), which are important positions in Mex-LanM23,26 and other EF-hand proteins29. Therefore, Hans-LanM presented an opportunity to determine features essential for selective lanthanide recognition in LanMs.

Hans-LanM was expressed in Escherichia coli as a 110 amino acid protein (Supplementary Fig. 1). LaIII and NdIII were selected as representative LREs and DyIII was selected as a representative HRE for complexation studies. The protein binds about three equivalents of LaIII and NdIII, and slightly less DyIII, by inductively coupled plasma mass spectrometry (Supplementary Table 1), as does Mex-LanM4. Also like Mex-LanM4, Hans-LanM exhibits little helical content in the absence of metal, as judged by the circular dichroism signal at 222 nm (Fig. 1c). Unexpectedly, only two equivalents of LaIII or DyIII were sufficient to cause Hans-LanM's complete conformational change (Supplementary Fig. 2), indicating that the third binding equivalent is weak and does not increase helicity.

The apparent dissociation constants (Kd,app) determined by circular dichroism spectroscopy4 reflect the RE versus RE, and RE versus non-RE, selectivities of Mex-LanM under competitive RE recovery conditions6,18. Therefore, similar determinations of Kd,app with free metal concentrations controlled by a competitive chelator4,32 were applied to Hans-LanM; the results (Fig. 1d and Supplementary Table 2) diverged from those for Mex-LanM. Binding of LaIII and NdIII to Hans-LanM increases molar ellipticity at 222 nm by 2.3-fold, the full conformational change evident in stoichiometric titrations. The conformational change is cooperative (Hill coefficients, n, of 2; Supplementary Table 2), and the Kd,app values are similar, 68 and 91 pM, respectively. By contrast, even though DyIII induces the same overall response as LaIII in stoichiometric titrations (Supplementary Fig. 2), in the chelator-buffered DyIII titrations Hans-LanM exhibits a lesser conformational response (1.8-fold increase). This difference indicates that at least one of the DyIII-binding sites is very weakly responsive (Kd,app > 0.3 µM, the highest concentration accessible in the chelator-buffered titrations). The main response to DyIII occurs at 2.6 nM, >30-fold higher than with the LREs, and with little or no cooperativity (n = 1.3). By contrast, Mex-LanM shows only a modest preference for LREs (about fivefold; Fig. 1e; ref. 4), and all lanthanides and YIII induce similar conformational changes and cooperativity18. Hans-LanM responds to calcium(II) weakly (Kd,app = 60 µM), with the same lack of cooperativity (n = 1.0) and partial conformational change evident with DyIII (Extended Data Fig. 2). Therefore, Hans-LanM discriminates more strongly between LREs and HREs than does Mex-LanM, with the HRE complexes exhibiting lower affinity, lesser cooperativity and a lesser primary conformational change.

The distinct behaviours of the LRE– and HRE–Hans-LanM complexes suggested mechanism(s) of LRE versus HRE selectivity not present in Mex-LanM. As Mex-LanM is a monomer in complex with LREs and HREs alike4,5, we considered that LREs and HREs might induce different oligomeric states in Hans-LanM. In the presence of three equivalents of LaIII, Hans-LanM elutes from a size-exclusion chromatography (SEC) column not at the expected molecular weight (MW) of 11.9 kDa but instead at 27.8 kDa, suggestive of a dimer (Supplementary Figs. 3 and 4a). Starting gradually after NdIII but sharply at GdIII, the apparent MW decreases towards that expected for a monomer (Fig. 2a, Supplementary Fig. 4 and Supplementary Table 3). Notably, lanthanides heavier than GdIII do not seem to support growth of RE-utilizing bacteria33,34,35.

a, Apparent molecular weight of Hans-LanM complexes with REs as determined by analytical SEC (red lines) or SEC–MALS (black dashed line). See Supplementary Table 1 for conditions. Each individual data point is the result of a single experiment. b, The LaIII-bound Hans-LanM dimer as determined by X-ray crystallography. LaIII ions are shown as green spheres and NaI ions are shown as grey spheres. c, Detailed view of the dimer interface near EF3 of chain A (blue cartoon). Arg100 from chain C (light blue cartoon) anchors a hydrogen-bonding network involving Asp93 of chain A and two EF3 LaIII ligands (Glu91 and Asp85). These interactions constitute the sole polar contacts at the dimer interface, providing a means to control the radius of the lanthanide-binding site at EF3. d, Schematic of the interactions at the dimer interface. Red dashed lines indicate hydrogen-bonding interactions and grey dashed lines indicate hydrophobic contacts. e, DENSS projections of electron density from small-angle X-ray scattering datasets for LaIII-bound (left) and DyIII-bound (right) Hans-LanM, overlaid with a PyMOL-generated ribbon diagram of the dimeric LaIII–Hans-LanM crystal structure.

To provide further support for preferential dimerization in the presence of physiologically relevant LREs, RE complexes of Hans-LanM were analysed using multi-angle light scattering (MALS; Fig. 2a and Supplementary Fig. 5). The LaIII, NdIII and GdIII complexes have MWs of 22–25 kDa, indicative of dimers, but MWs decrease starting with TbIII and continue to DyIII and HoIII, at about 15 kDa (Extended Data Table 1), in agreement with the SEC data. CaII-bound Hans-LanM also indicated a MW of 14.7 kDa. The HRE–, CaII– and apo Hans-LanM complexes are still one-third larger than expected for a monomer, however, suggesting that these forms exist in a rapid equilibrium with ≈2:1 monomer/dimer ratio under these conditions. We next determined the Kd for dimerization (Kdimer) of apo, LaIII-bound and DyIII-bound Hans-LanM by isothermal titration calorimetry (Extended Data Table 2 and Supplementary Figs. 6–8). The apoprotein and DyIII-bound protein weakly dimerize, with Kdimer values of 117 µM and 60 µM, respectively, consistent with the ratios of monomer and dimer reflected in the SEC and MALS traces. In the presence of LaIII, however, the dimer was too tight to be able to observe monomerization by isothermal titration calorimetry, which indicates that Kdimer <0.4 µM (Supplementary Fig. 8). Thus, LaIII favours Hans-LanM's dimerization by >100-fold over DyIII.

A 1.8-Å-resolution X-ray crystal structure of Hans-LanM in complex with LaIII confirms LRE-induced dimerization (Extended Data Fig. 3 and Supplementary Table 4). Two LanM monomers interact head-to-tail (Fig. 2b), burying about 600 Å2 of surface area through hydrophobic and polar contacts (Fig. 2c,d). These interactions occur largely between side chains contributed by the core helices α1 (between EF1 and EF2) and α2 (between EF3 and EF4; Supplementary Fig. 9). Residues at the dimer interface make direct contact with only one of the four metal-binding sites, EF3; three residues of EF3 in each monomer form a hydrogen-bonding network with Arg100 of the other monomer (Fig. 2c), suggesting that occupancy and coordination geometry at this site may control oligomeric state.

Hans-LanM and its complexes with three equivalents of LaIII, NdIII and DyIII were also analysed by small-angle X-ray scattering (Supplementary Figs. 10 and 11). The calculated solvent envelopes36 from the small-angle X-ray scattering data fit well to the crystallographic Hans-LanM dimer for LaIII–Hans-LanM, adequately for NdIII–Hans-LanM, but poorly for DyIII–Hans-LanM (Fig. 2e and Supplementary Figs. 12–14). The weaker dimerization of DyIII–Hans-LanM is also supported by quantitative metrics, such as the Porod volume (Supplementary Figs. 15 and 16 and Supplementary Tables 5 and 6). Together, the biochemical and structural results indicate that Hans-LanM's dimerization equilibrium depends strongly on the particular RE bound.

The structure of LaIII–Hans-LanM also provides one of the first detailed views of the coordination environments in a LanM, and indeed any natural biomolecule tasked with reversible lanthanide recognition. The previous NMR structure of Mex-LanM5 revealed the protein's unusual fold, but it could not provide molecular-level details about the metal-binding sites. To understand the basis for LRE versus HRE discrimination, we also determined a 1.4-Å-resolution structure of DyIII–Hans-LanM. Finally, we report a 1.01-Å-resolution structure of NdIII–Mex-LanM, which rationalizes Mex-LanM's shallower RE selectivity trend4.

In LaIII–Hans-LanM, EF1–3 are occupied by LaIII ions (Extended Data Fig. 3b–e). EF4 is structurally distinct, does not exhibit anomalous difference density consistent with LaIII and was modelled with NaI (Supplementary Fig. 17a). Each LaIII-binding site is ten-coordinate, as observed in structures of lanthanide-dependent methanol dehydrogenases33,37 (Supplementary Fig. 18). A monodentate Asn (N1 position), four bidentate Asp or Glu residues (D3, D5, E9 and E12) and a backbone carbonyl (T7 or S7) complete the first coordination sphere in EF1–3 (Fig. 3a). Exogenous solvent ligands are not observed (Supplementary Fig. 17b); luminescence studies of EuIII–Hans-LanM to determine the number of coordinated solvent molecules (q) yielded q = 0.11, consistent with the absence of solvent ligands in the X-ray structure (Supplementary Fig. 19).

a, Zoomed-in views of EF2 (left) and EF3 (right) in LaIII–Hans-LanM. LaIII ions are shown as green spheres. Coordination bonds and hydrogen bonds are shown as dashed lines. Residues contributed by chain A are shown in blue and those contributed by chain C (in the case of EF3) are shown in light blue. Inset: overlay of LaIII–Hans-LanM (blue and light blue) with DyIII–Hans-LanM (grey), showing the carboxylate shift of Glu91 from bidentate (La) to monodentate (Dy). Coordination and hydrogen bonds (dashed lines) are shown only for the Dy case. b, Representative metal-binding site (EF3) in NdIII–Mex-LanM. NdIII ion is shown as an aqua sphere. Solvent molecules are shown as red spheres.

The lanthanide-binding sites in Hans-LanM additionally share extensive second-sphere interactions that may further constrain the positions of the ligands and the size of the metal-binding pore (Supplementary Fig. 20). This phenomenon is most obvious in EF3, at which the dimer interface mediates an extended hydrogen-bonding network involving several ligands. Arg100, contributed by the adjacent monomer, projects into the solvent-exposed side of EF3 to contact two carboxylate ligands, Asp85 (D3) and Glu91 (E9), enforcing their bidentate binding modes. Arg100 is also buttressed by Asp93 (EF3 D11), unique to EF3 within Hans-LanM and not observed in Mex-LanM. We tested the importance of this network in Hans-LanM dimerization by making the minimal substitution, R100K. Hans-LanM(R100K) had nearly identical Kd,app values and response to NdIII and DyIII as wild-type Hans-LanM, but the Kd,app for LaIII was twofold weaker (Supplementary Fig. 21 and Supplementary Table 7). SEC–MALS analysis indicated MWs of 10–13 kDa for apo, LaIII– and DyIII–Hans-LanM(R100K) (Supplementary Fig. 22 and Supplementary Table 8), indicative of increased monomerization, especially for the LaIII complex, and suggesting that weaker dimerization may be responsible for the lower LaIII affinity. All four residues comprising the Arg100–EF3 network are completely conserved in the Hans cluster (Supplementary Fig. 23), suggesting that these interactions may contribute to dimerization in these LanMs.

The structure of DyIII–Hans-LanM confirms the importance of second-sphere control of ligand positioning (Extended Data Fig. 4, Supplementary Figs. 24–26 and Supplementary Tables 9 and 10). The overall structure of DyIII–Hans-LanM is largely superimposable with that of LaIII–Hans-LanM, and the coordination spheres of the DyIII ions in EF1–3 are similar to those in LaIII–Hans-LanM (Fig. 3a, inset), with the notable exception of E9 (for example, Glu91 in EF3). This residue shifts from bidentate with LaIII to monodentate with the smaller DyIII ions, yielding a nine-coordinate distorted capped square antiprismatic geometry; the lower coordination number with a HRE ion is consistent with the case of other RE complexes38,39. In EF3, this carboxylate shift lengthens the distance between Arg100 and the proximal Oε of Glu91 from 2.9 Å (in LaIII–Hans-LanM) to 3.2 Å (Supplementary Fig. 27). The rearrangement of this second-sphere hydrogen-bonding network suggests a structural basis for RE-dependent differences in Kdimer values.

The metal-binding sites of Mex-LanM differ substantially from those of Hans-LanM. In Mex-LanM, all four EF hands are occupied by nine-coordinate (EF1–3) or ten-coordinate (EF4) NdIII ions, each including two solvent ligands, not present in Hans-LanM (Fig. 3b and Supplementary Fig. 28). The observation of the two solvent molecules per metal site and the hydrogen bond to the D9 residue validates recent spectroscopic studies21,23,26. The difference in coordination number between EF1–3 and EF4 is due to the D3 carboxylate being monodentate in EF1–3 but bidentate in EF4. Although the NdIII sites of Mex-LanM share the nine- and ten-coordination observed in DyIII– and LaIII–Hans-LanM, they structurally resemble the seven-coordinate CaII-binding sites of calmodulin (Supplementary Fig. 18). The increased coordination numbers in Mex-LanM relative to calmodulin result from bidentate coordination of D5 and an additional solvent ligand. These similarities suggest that much of LanM's unique 108-fold selectivity for REs over CaII results from subtle differences in second-coordination-sphere and other more distal interactions. Finally, the exclusively protein-derived first coordination sphere in Hans-LanM, particularly due to coordination by E9, yields more extended hydrogen-bonding networks (Supplementary Figs. 20 and 29) and probably enhances control over the radius of the binding site. Thus, the structures rationalize the extraordinary RE versus non-RE selectivity of Mex-LanM and Hans-LanM while also accounting for their differences in LRE versus HRE selectivity.

The differences in stability and structure between Hans-LanM's LRE versus HRE complexes suggested that Hans-LanM (wild type and/or R100K) would outperform Mex-LanM in RE/RE separations. We focused on separating the RE pair of NdIII and DyIII used in permanent magnets. We first assayed the stabilities of the wild-type Hans-LanM and Hans-LanM(R100K) RE complexes against citrate, previously used as a desorbent with Mex-LanM6. RE–Hans-LanM complexes are generally less stable against citrate than those of Mex-LanM, as expected on the basis of lower affinity (Fig. 1e), but the difference in stability between the NdIII–Hans-LanM and DyIII–Hans-LanM complexes—expressed as the ratio of citrate concentration required for 50% desorption of each metal ([citrate]1/2), as reported by the fluorescence of Hans-LanM's two Trp residues (Supplementary Fig. 30)—is twofold greater than for Mex-LanM complexes (Fig. 4a, Supplementary Table 11 and Extended Data Fig. 5). Furthermore, the R100K substitution significantly destabilizes Hans-LanM's LaIII complex against citrate, whereas it only slightly affects the NdIII complex and does not affect the DyIII complex. This result confirms that dimerization selectively stabilizes Hans-LanM's LRE complexes (and especially the LaIII complex), a factor abrogated by the R100K substitution. Using malonate, a weaker chelator than citrate, DyIII can be readily desorbed from both Hans-LanM and R100K with 10–100 mM chelator without significant NdIII desorption, suggesting conditions for NdIII/DyIII separation (Fig. 4b).

a, Hans-LanM and the R100K variant exhibit greater differences in Nd versus Dy complex stability than Mex-LanM against desorption by citrate. Mean ± s.e.m. for three independent trials. **Significant difference between [citrate]1/2 for LaIII between Hans-LanM and Hans-LanM(R100K) (20 µM protein) shows the impact of dimerization of LaIII complex stability (P < 0.01, analysis of variance with Bonferroni post-test). Mex-LanM Nd and Dy data from ref. 6. b, Spectrofluorometric titration of Hans-LanM and R100K variant (λex = 280 nm, λem = 333 nm) at pH 5.0, depicting the malonate-induced desorption of a 2:1 metal–protein complex. Mean ± s.e.m. for three independent trials, except those with R100K, which were single trials of each condition. c, Comparison of distribution factors (pH 5.0, about 0.33 mM each RE, LaIII–DyIII) for immobilized Hans-LanM, Hans-LanM(R100K) and Mex-LanM. Each point represents mean ± s.d. for three independent trials. d, Separation of a 95:5 mixture of NdIII/DyIII using immobilized Hans-LanM(R100K) and a desorption scheme of three stepped concentrations of malonate followed by pH 1.5 HCl. One bed volume was 0.7 ml.

Source Data

Although a twofold modulation of RE versus RE selectivity by dimerization may seem small, such differences provide opportunity to decrease the number of separation stages, increasing efficiency of a separation process3,12. Therefore, Hans-LanM and the R100K variant were immobilized through a carboxy-terminal Cys residue on maleimide-functionalized agarose beads, as described previously6, and tested for NdIII/DyIII separation. Immobilized Hans-LanM bound about one equivalent of RE, unlike in solution and compared with two equivalents for Mex-LanM6 and Hans-LanM(R100K) (Supplementary Fig. 31). Hans-LanM and R100K exhibited similar separation ability in the La–Gd range—although R100K exhibits greater separation ability in the Gd–Dy range—as determined by the on-column distribution ratios (D) of a mixed RE solution at equilibrium (Fig. 4c, Extended Data Table 3 and Supplementary Tables 12–14). These Nd/Dy separation factors are nearly double (Hans-LanM) and triple (Hans-LanM(R100K)) that of Mex-LanM (Extended Data Table 3). Immobilized Hans-LanM was loaded to 90% of breakthrough capacity with a model electronic waste mixture of 5% dysprosium and 95% neodymium and, guided by Fig. 4b, eluted with a short, stepwise malonate gradient, followed by complete desorption using pH 1.5 HCl. In a single purification stage, Dy was upgraded from 5% to 83% purity and Nd was recovered at 99.8% purity (both >98% yield; Extended Data Fig. 6). This significantly outperformed the comparable Mex-LanM-based process, which achieved only 50% purity in a first separation stage and required a second stage to obtain >98% purity6. The immobilized R100K variant performed even better, achieving baseline separation of DyIII and NdIII to >98% purity and >99% yield in a single stage (Fig. 4d). The R100K variant's better performance was unexpected and may point to the unlikelihood of functional dimers on the column at this immobilization density (see the caption of Extended Data Fig. 6 for a discussion). Thus, despite substantially improved performance versus Mex-LanM enabled by characterization of Hans-LanM's mechanism of dimerization, fully exploiting the dimerization phenomenon on-column may involve, for example, tethering of two monomers on a single polypeptide chain, which is under investigation.

Biochemical and structural characterization of Hans-LanM's mechanism of metal-sensitive dimerization provides a new, allosteric mechanism for LRE versus HRE selectivity in biology, extending concepts in dimer-dependent metal recognition recently emerging from synthetic lanthanide complexes11 and engineered transition metal-binding proteins40 and showing that these principles are hard-wired into nature. Our work also shows that dimerization strength, and thus metal selectivity, can be rationally modulated. Hans-LanM evolved LRE-selective dimerization at physiological protein concentrations closer to those in our biochemical assays (10–20 µM) rather than those on the column (about 3 mM). Therefore, leveraging dimerization in a separation process would be assisted by shifting dimerization sensitivity to the higher concentration regime, such as by tuning hydrophobic interactions at the dimerization interface. Furthermore, our studies establish that LanMs with as low as 33% identity are easily predicted yet have useful differences in metal selectivity; further mining of this diversity may reveal yet additional mechanisms for tuning RE separations. Finally, the solvent-excluded coordination spheres of Hans-LanM should outperform Mex-LanM in RE/actinide separation23, luminescence-based sensing21,26 and stabilization of hydrolysis-prone ions. Continued characterization of the coordination and supramolecular principles of biological f-element recognition will inspire design of ligands with higher RE versus RE selectivities and their implementation in new RE separation processes.

See the Supplementary Methods for details.

The sequence of LanM from M. extorquens AM1 was used as a query to conduct PSI-BLAST searches against the National Center for Biotechnology Information non-redundant protein sequence (nr) and metagenomic protein (env_nr) databases until convergence41. The resulting 3,047 protein sequences were then manually curated for those that are less than 200 residues long, have at least one pair of EF hands separated by less than 14 residues, and have 4 EF hands. Signal peptides of LanM sequences were predicted using SignalP (v6.0)42, and then removed before further analysis of the sequences.

The Enzyme Function Initiative-Enzyme Similarity Tool was used to calculate the similarities between all peptide sequence pairs with an E-value threshold of 1 × 10−5 (ref. 30). The resulting sequence similarity network of 696 nodes and 241,853 edges was then constructed and explored using the organic layout through Cytoscape (v3.9.1)43 and visualized in R (v4.1.0)44. The edge percentage identity threshold was gradually increased from 40% to 90% to yield distinct clusters.

LanM sequences were aligned using MUSCLE (v5.1)45 with default parameters. The model used for phylogeny construction was selected using ModelFinder in IQ-TREE (v2.2.0.3)46,47 with --mset set to beast2. Bayesian phylogeny was generated on the basis of these results using BEAST (v2.6.7)48. The resulting phylogeny was evaluated using 107 generations and discarding a burn-in of 25%, and then visualized using ggtree (v3.2.1)49.

The gene encoding Hans-LanM, codon optimized for expression in E. coli without its native 23-residue signal peptide (see Supplementary Table 15), was obtained from Twist Bioscience and inserted into pET-29b(+) using the restriction sites NdeI/XhoI. Hans-LanM was overexpressed on a 2-l scale and purified using the established protocol for Mex-LanM50, with one modification: after the final SEC step, the protein was concentrated to 5 ml and dialysed against 5 g Chelex 100 in 500 ml of 30 mM HEPES, 100 mM KCl, 5% glycerol, pH 8.4, to remove CaII and trace metal contaminants. This procedure resulted in approximately 15 ml of 550 μM protein, which was not concentrated further. The final yield was 45 mg of protein per litre of culture. Protein concentrations were calculated using an extinction coefficient of 11,000 M−1 cm−1, based on the ExPASy ProtParam tool51. Hans-LanM(R100K) was purified using the same procedure, yielding 30 mg of protein per litre of culture.

Circular dichroism spectra of Hans-LanM were collected as described previously32, at 15 µM (monomer concentration) in Chelex 100-treated buffer A (20 mM acetate, 100 mM KCl, pH 5.0), unless otherwise indicated. Buffered metal solutions were prepared as described previously4,23,25,32. Additional details are available in the Supplementary Information.

Samples of wild-type Hans-LanM were prepared by adding 3.0 equivalents of metal slowly (0.5 equivalent at a time followed by mixing) to 1.0 ml of concentrated stock of Hans-LanM (550 μM). At these protein concentrations, slight precipitation was observed for LRE samples (for example, LaIII) whereas significant precipitation was observed for HRE samples (for example, DyIII). Samples were centrifuged at 12,000g for 2 min to remove precipitate and then purified using gel filtration chromatography (HiLoad 10/300 Superdex 75 pg, 1-ml loop, 0.8 ml min−1) in buffer B (30 mM MOPS, 100 mM KCl, 5% glycerol, pH 7.0). Hans-LanM-containing peaks (ranging from 12.0 to 15.0 ml elution volume) were collected to avoid the high-MW aggregate peaks, yielding 2.0 ml of metalated Hans-LanM ranging between 114 μM and 128 μM (1.37–1.53 mg ml−1).

Samples of Hans-LanM(R100K) do not form high-MW species or precipitate on metal addition. To prepare samples of this protein, a 500 μM protein solution was diluted to 250 μM (3 mg ml−1) in buffer B containing 0.75 mM of a specific RECl3, yielding a final solution of 3 mg ml−1 protein, with a 1:3 metal ratio, which was analysed directly by SEC–MALS.

For calcium conditions, proteins were diluted to 250 μM (3 mg ml−1), 5 mM CaCl2 was added, and the samples were incubated at room temperature for 1 h. The buffer used for SEC–MALS was the same as above, except that it also contained 5 mM CaCl2.

SEC–MALS experiments were conducted using an Agilent 1260 Infinity II HPLC system equipped with an autosampler and fraction collector, and the Wyatt SEC hydrophilic column had 5-µm silica beads, a pore size of 100 Å and dimensions of 7.8 × 300 mm. Wyatt Technology DAWN MALS and Wyatt Optilab refractive index detectors were used for analysing the molar mass of peaks that eluted from the column. The SEC–MALS system was equilibrated for 5 h with buffer B. The system was calibrated with bovine serum albumin (monomer MW: 66 kDa) in the same buffer and normalization and alignment of the MALS and refractive index detectors were carried out. A volume of 15 µl of each sample was injected at a flow rate of 0.8 ml min−1 with a chromatogram run time of 25 min. Data were analysed using the ASTRA software (Wyatt). When small-angle X-ray scattering (SAXS) analysis was desired, a second run was carried out with 150 µl protein (about 4 mg ml−1) injected, and 200-µl fractions of the main peak were collected. BioSAXS data were subsequently collected in triplicate.

The dissociation constants for the dimers of apo, LaIII-bound and DyIII-bound Hans-LanM were determined by dilutive additions of a concentrated protein stock, followed using isothermal titration calorimetry on a TA Instruments Low-volume Auto Affinity isothermal titration calorimeter. The syringe contained 300 μM protein (apo or 2 equivalents of DyIII bound) or 150 µM or 540 µM (2 equivalents of LaIII bound), and the cell contained 185 μl of a matched buffer (30 mM MOPS, 100 mM KCl, pH 7.0). Titrations were carried out at 30 °C. Titrations consisted of a first 0.2-μl injection followed by 17 × 2-μl injections, unless otherwise noted, with stirring at 125 r.p.m. and 180 s equilibration time between injections. The data were fitted using NanoAnalyze using the Dimer Dissociation model, yielding the dimer dissociation constant (Kdimer), enthalpy of dissociation (ΔH) and entropy of dissociation (ΔS). All parameters are shown in Extended Data Table 2.

Kdimer is defined as the dissociation constant for the equilibrium D \(\rightleftharpoons \) 2M, such that Kdimer = [M]2/[D], in which [D] is the concentration of the dimer and [M] is the concentration of the monomer, and the total protein concentration [P] (as measured using the extinction coefficient for the monomer) is given by [P] = [M] + 2[D]. Therefore, Kdimer = 2[M]2/([P] − [M]) or

This equation can be used to estimate monomer and dimer concentrations during SEC–MALS experiments, using Kdimer values calculated from isothermal titration calorimetry experiments and [P] from the SEC–MALS trace. This equation can also be used to estimate the maximum possible Kdimer for LaIII-bound protein, given the SEC–MALS data.

SAXS data were collected on RE-complexed Hans-LanM, at protein concentrations given in Supplementary Table 5 using equipment and under conditions described in the Supplementary Methods.

The forward scattering I(0) and the radius of gyration (Rg) are listed in Supplementary Table 5 and were calculated using the Guinier approximation, which assumes that at very small angles (q < 1.3/Rg) the intensity is approximated as I(q) =  I(0)exp[−1/3(qRg)2]. In the LaIII-, NdIII- and DyIII-bound conditions, this agrees with the calculated size of 17.9 Å for the crystallographic dimer. The molecular mass was estimated using a comparison with SAXS data of a bovine serum albumin standard. The data files were analysed for Guinier Rg, maximum particle dimension (Dmax), Guinier fits, Kratky plots and pair-distance distribution function using the ATSAS software52. GNOM, within ATSAS, was used to calculate the pair-distance distribution function P(r), from which Rg and Dmaxwere determined. Solvent envelopes were computed using DENSS36. The theoretical scattering profiles of the constructed models were calculated and fitted to experimental scattering data using CRYSOL53. OLIGOMER54 was used to estimate the monomer and dimer fractions.

To Hans-LanM (2 ml, 1.16 mM, buffer B), 3.0 equivalents of LaCl3 or DyCl3 were added slowly, 0.5 equivalents at a time with mixing, to minimize precipitation. Precipitate was removed by centrifugation at 12,000g for 2 min. Any soluble aggregates were removed and the protein was exchanged into buffer lacking glycerol (buffer C: 30 mM MOPS, 50 mM KCl, pH 7.0) by gel filtration chromatography (HiLoad 16/600 Superdex 75 pg, 1-ml loop, 0.75 ml min−1). The peak in the 70–85 ml range was pooled, and the fractions were concentrated to about 500 μl with a final concentration of about 1.3 mM.

Mex-LanM was purified as described previously50 and was exchanged into buffer C before crystallization. The protein was loaded with 3.5 equivalents of NdIII (NdCl3).

Diffraction datasets were collected at the Life Sciences Collaborative Access Team ID-G beamline and processed with the HKL2000 package55. In all structures, phase information was obtained with phenix.autosol56,57 through the single-wavelength anomalous diffraction method, in which lanthanide ions identified with HySS58 were used as the anomalous scatterers. Initial models were generated with phenix.autobuild59 with subsequent rounds of manual modification and refinement in Coot60 and phenix.refine61. In the final stages of model refinement, anisotropic displacement parameters and occupancies were refined for all lanthanide sites62. Model validation was carried out with the Molprobity server63. Figures were prepared using the PyMOL molecular graphics software package (Schrödinger, LLC).

Crystals were obtained by using the sitting drop vapour diffusion method, in which 1 μl of protein solution (15 mg ml−1) was mixed with 1 μl 10 mM tri-sodium citrate, pH 7.0, and 27% (w/v) PEG 6000 in a 24-well plate from Hampton Research (catalogue number HR1-002) at room temperature. Thin plate-shaped crystals appeared in 3 days. Crystals suitable for data collection were mounted on rayon loops, soaked briefly in a cryoprotectant solution consisting of the well solution supplemented with 10% ethylene glycol, and flash-frozen in liquid N2.

LaIII-loaded Hans-LanM crystallized in the P21 space group (β = 90.024°) with four monomers in the asymmetric unit. The initial figure of merit and Bayesian correlation coefficient were 0.563 and 0.56, respectively64. The final model consists of residues 24–133 in each chain, 12 LaIII ions (3 per chain in the first, second and third EF hands), 4 NaI ions65 (1 per chain in the fourth EF hand), 273 water molecules and 2 molecules of citrate. Of the residues modelled, 100% are in allowed or preferred regions as indicated by Ramachandran statistical analysis.

Crystals were obtained by using the sitting drop vapour diffusion method, in which 1 μl of protein solution (15 mg ml−1) was mixed with 1 μl of 250 μM tri-sodium citrate, pH 7.0, and 27% (w/v) PEG 6000 in a 24-well plate from Hampton Research at room temperature. Thin plate-shaped crystals appeared within 1 month. Crystals suitable for data collection were mounted on rayon loops, soaked briefly in a cryoprotectant solution consisting of the well solution supplemented with perfluoropolyether cryo oil from Hampton Research (catalogue number HR2-814) and flash-frozen in liquid N2.

DyIII-loaded Hans-LanM crystallized in the P21 space group (β = 93.567°) with four monomers in the asymmetric unit. The initial figure of merit and Bayesian correlation coefficient were 0.748 and 0.58, respectively64. The final model consists of residues 24–133 in each chain (except for chain D, for which residues 34–38 cannot be modelled), 14 DyIII ions (4 in chains A and D, 3 in the second, third and fourth EF hands of chains B and C) and 656 water molecules. Of the residues modelled, 100% are in allowed or preferred regions as indicated by Ramachandran statistical analysis.

Collection of anomalous datasets is described in the Supplementary Methods.

Crystals were obtained by using the sitting drop vapor diffusion method, in which 1 μl of protein solution (35 mg ml−1) was mixed with 1 μl of 0.1 M ammonium sulfate, 0.1 M Tris pH 7.5, and 20% (w/v) PEG 1500 in a 24-well plate from Hampton Research at room temperature. Thin plate-shaped crystals appeared within 6 months. Crystals suitable for data collection were mounted on rayon loops, soaked briefly in a cryoprotectant solution consisting of the well solution supplemented with perfluoropolyether cryo oil from Hampton Research and flash-frozen in liquid N2.

NdIII-loaded Mex-LanM crystallized in the P212121 space group with one monomer in the asymmetric unit. The initial figure of merit and Bayesian correlation coefficient were 0.799 and 0.56, respectively64. The final model consists of residues 29–133, 4 NdIII ions and 171 water molecules. Of the residues modelled, 100% are in allowed or preferred regions as indicated by Ramachandran statistical analysis.

All fluorescence data were collected using a Fluorolog-QM fluorometer in configuration 75-21-C (Horiba Scientific) equipped with a double monochromator on the excitation arm and single monochromator on the emission arm. A 75-W xenon lamp was used as the light source for steady-state measurements and a pulsed xenon lamp was used for time-resolved measurements. Ten-millimetre quartz spectrofluorometry cuvettes (Starna Cells, 18F-Q-10-GL14-S) were used to collect data at 90° relative to the excitation path.

Fluorescence lifetime measurements were carried out using established methods26,66. In short, a solution of Hans-LanM with 2 equivalents of EuIII added, totalling 4.5 ml, was prepared in 100% H2O matrix (buffer: 25 mM HEPES, 75 mM KCl, pH 7.0). Half of this initial protein mixture (2.25 ml) was retained for future use and the remainder was exchanged to D2O through lyophilization to remove H2O and resuspension in 99.9% D2O two times. The resulting protein solutions (in 100% H2O and about 99% D2O) were mixed in varying ratios to produce D2O contents of 0%, 25%, 50% and 75%. The protein concentration was 20 µM. For each sample, the luminescence decay time constant (τ) was measured (λex = 394 nm, λem = 615 nm) with 5,000 shots over a time span of 2,500 μs. τ was determined using the FelixFL Powerfit-10 software (Horiba Scientific) using a single exponential fit. 1/τ was plotted against percentage composition of D2O, and the slope of the resulting line (m) was determined. The q value was determined using the following equation from ref. 67:

in which τ−1H2O and τ−1D2O are the inverses of the time constants in 100% H2O and D2O, respectively (the latter extrapolated using the equation of the fitted line), in ms–1; and nOH = 0, nNH = 0, and nO–CNH = 1 (resulting from the metal-coordinated Asn residues), on the basis of the Hans-LanM crystal structures. This equation simplifies to:

For fluorescence competition experiments, a solution of 20 μM Hans-LanM or the R100K variant was prepared in buffer A (pH 5.0) with two equivalents of metal (40 μM). Fluorescence emission spectra were collected with settings: λex = 278 nm, λem 300–420 nm, integration time = 0.5 s, step size = 1 nm. Titrations were carried out through addition of at least 0.6 μl of titrant (from concentrated stock solutions of 10 mM–1 M citrate or malonate, pH 5.0). Spectra were corrected for dilution. Each experiment was carried out in triplicate.

Hans-LanM(R100K)-Cys was expressed and purified as described for Mex-LanM-Cys (ref. 6), with a final yield of 50 mg of protein per litre of culture. For Hans-LanM-Cys, the protein was purified by incorporating the same modifications from above, minus the dialysis step, to our previously described Mex-LanM-Cys purification, except that the SEC step was run using a reducing buffer (30 mM MOPS, 100 mM KCl, 5 mM TCEP, pH 7.0) with 5 mM EDTA, and frozen under liquid N2 before immobilization.

The maleimide functionalization of amine-functionalized agarose beads was described previously6. See the Supplementary Information for complete details.

Hans-LanM(R100K) immobilization was carried out using a thiol-maleimide conjugation reaction as described previously6. In the case of Hans-LanM, a final protein concentration of about 0.4 mM (8 ml) was combined with 1 ml of maleimide–microbeads and the conjugation reaction was carried out for 16 h at room temperature. Unconjugated Hans-LanM was removed by washing with coupling buffer, and the Hans-LanM microbeads were stored in coupling buffer for subsequent tests. To quantify Hans-LanM immobilization yield, Pierce BCA Protein Assay (ThermoFisher Scientific) was used to determine the LanM concentration in the reaction solution before and after the conjugation reaction as previously described.

LanM-immobilized microbeads were washed with deionized water. Feed solution (5 ml, equimolar REs La–Dy, 3 mM total, pH 5.0) was added to 1 ml microbeads and incubated for 2 h. The liquid at equilibrium was collected and RE concentrations were determined by inductively coupled plasma mass spectrometry as [M]ad. Then 4 ml of 0.1 M HCl was used to desorb REs from the microbeads and concentrations were measured by inductively coupled plasma mass spectrometry as [M]de.

The RE distribution factor (D) between the LanM phase and the solution phase was calculated as:

in which [M]LanM and [M]Liquid are the molar concentrations of each metal ion in the LanM phase and the solution phase at equilibrium, respectively. To account for the free liquid that absorbed on the agarose microbeads, the following correction was applied: [M]Liquid = [M]ad; [M]LanM = (4 × [M]de – [M]ad)/4.

The separation factor is defined as:

in which DRE1 and DRE2 are the distribution factors of RE1 and RE2, respectively.

Columns were filled and run, and metal concentrations analysed, as described in our previous work6; details are available in the Supplementary Methods.

For the RE pair separation experiments, the metal ion purity and yield are defined as:

in which CRE1 and CRE2 are the molar concentrations of RE1 and RE2, respectively.

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

All data are available in the main text or the Supplementary Information. Coordinates have been deposited in the Protein Data Bank with accession codes 8DQ2 (LaIII–Hans-LanM), 8FNR (DyII–Hans-LanM) and 8FNS (NdIII–Mex-LanM). Source data are provided with this paper.

Cheisson, T. & Schelter, E. J. Rare earth elements: Mendeleev's bane, modern marvels. Science 363, 489–493 (2019).

Article ADS CAS PubMed Google Scholar

Nash, K. L. & Jensen, M. P. Analytical-scale separations of the lanthanides: a review of techniques and fundamentals. Sep. Sci. Technol. 36, 1257–1282 (2001).

Article CAS Google Scholar

Xie, F., Zhang, T. A., Dreisinger, D. & Doyle, F. A critical review on solvent extraction of rare earths from aqueous solutions. Miner. Eng. 56, 10–28 (2014).

Article CAS Google Scholar

Cotruvo, J. A. Jr, Featherston, E. R., Mattocks, J. A., Ho, J. V. & Laremore, T. N. Lanmodulin: a highly selective lanthanide-binding protein from a lanthanide-utilizing bacterium. J. Am. Chem. Soc. 140, 15056–15061 (2018).

Article CAS PubMed Google Scholar

Cook, E. C., Featherston, E. R., Showalter, S. A. & Cotruvo, J. A. Jr. Structural basis for rare earth element recognition by Methylobacterium extorquens lanmodulin. Biochemistry 58, 120–125 (2019).

Article CAS PubMed Google Scholar

Dong, Z. et al. Bridging hydrometallurgy and biochemistry: a protein-based process for recovery and separation of rare earth elements. ACS Cent. Sci. 7, 1798–1808 (2021).

Article CAS PubMed PubMed Central Google Scholar

Shannon, R. D. Revised effective ionic radii and systematic studies of interatomic distances in halides and chalcogenides. Acta Cryst. A 32, 751–767 (1976).

Article Google Scholar

Deblonde, G. J. P., Chagnes, A. & Cote, G. Recent advances in the chemistry of hydrometallurgical methods. Sep. Purif. Rev. https://doi.org/10.1080/15422119.2022.2088389 (2022).

Vahidi, E. & Zhao, F. Environmental life cycle assessment on the separation of rare earth oxides through solvent extraction. J. Environ. Manage. 203, 255–263 (2017).

Article CAS PubMed Google Scholar

Jensen, M. P. et al. Solvent extraction separation of trivalent americium from curium and the lanthanides. Solvent Extr. Ion Exch. 33, 329–345 (2015).

Article CAS Google Scholar

Bogart, J. A. et al. Accomplishing simple, solubility-based separations of rare earth elements with complexes bearing size-sensitive molecular apertures. Proc. Natl Acad. Sci. USA 113, 14887–14892 (2016).

Article ADS CAS PubMed PubMed Central Google Scholar

Healy, M. R. et al. Efficient separation of light lanthanides(III) by using bis-lactam phenanthroline ligands. Chem. Eur. J. 25, 6326–6331 (2019).

Article CAS PubMed Google Scholar

Thiele, N. A., Fiszbein, D. J., Woods, J. J. & Wilson, J. J. Tuning the separation of light lanthanides using a reverse-size selective aqueous complexant. Inorg. Chem. 59, 16522–16530 (2020).

Article CAS PubMed Google Scholar

Slope, L. N., Daubney, O. J., Campbell, H., White, S. A. & Peacock, A. F. A. Location-dependent lanthanide selectivity engineered into structurally characterized designed coiled coils. Angew. Chem. Int. Ed. 60, 24473–24477 (2021).

Article CAS Google Scholar

Fang, H. et al. Electro-kinetic separation of rare earth elements using a redox-active ligand. Angew. Chem. Int. Ed. 56, 13450–13454 (2017).

Article ADS CAS Google Scholar

Demir, S. et al. Extraction of lanthanide and actinide ions from aqueous mixtures using a carboxylic acid-functionalized porous aromatic framework. ACS Cent. Sci. 2, 253–265 (2016).

Article CAS PubMed PubMed Central Google Scholar

Park, D. M. et al. Bioadsorption of rare earth elements through cell surface display of lanthanide binding tags. Environ. Sci. Technol. 50, 2735–2742 (2016).

Article ADS CAS PubMed PubMed Central Google Scholar

Deblonde, G. J.-P. et al. Selective and efficient biomacromolecular extraction of rare-earth elements using lanmodulin. Inorg. Chem. 59, 11855–11867 (2020).

Article CAS PubMed Google Scholar

Lumpe, H. et al. The earlier the better: structural analysis and separation of lanthanides with pyrroloquinoline quinone. Chem. Eur. J. 26, 10133–10139 (2020).

Article CAS PubMed Google Scholar

Ding, Y., Harvey, D. & Wang, N.-H. L. Two-zone ligand-assisted displacement chromatography for producing high-purity praseodymium, neodymium, and dysprosium with high yield and high productivity from crude mixtures derived from waste magnets. Green Chem. 22, 3769–3783 (2020).

Article CAS Google Scholar

Deblonde, G. J. P. et al. Characterization of americium and curium complexes with the protein lanmodulin: a potential macromolecular mechanism for actinide mobility in the environment. J. Am. Chem. Soc. 143, 15769–15783 (2021).

Article CAS PubMed Google Scholar

Deblonde, G. J.-P. et al. Capturing an elusive but critical element: natural protein enables actinium chemistry. Sci. Adv. 7, eabk0273 (2021).

Article ADS CAS PubMed PubMed Central Google Scholar

Mattocks, J. A., Cotruvo, J. A. Jr. & Deblonde, G. J. P. Engineering lanmodulin's selectivity for actinides over lanthanides by controlling solvent coordination and second-sphere interactions. Chem. Sci. 13, 6054–6066 (2022).

Article CAS PubMed PubMed Central Google Scholar

Singer, H., Drobot, B., Zeymer, C., Steudtner, R. & Daumann, L. J. Americium preferred: lanmodulin, a natural lanthanide-binding protein favors an actinide over lanthanides. Chem. Sci. 12, 15581–15587 (2021).

Article CAS PubMed PubMed Central Google Scholar

Mattocks, J. A., Ho, J. V. & Cotruvo, J. A. Jr. A selective, protein-based fluorescent sensor with picomolar affinity for rare earth elements. J. Am. Chem. Soc. 141, 2857–2861 (2019).

Article CAS PubMed Google Scholar

Featherston, E. R., Issertell, E. J. & Cotruvo, J. A. Jr. Probing lanmodulin's lanthanide recognition via sensitized luminescence yields a platform for quantification of terbium in acid mine drainage. J. Am. Chem. Soc. 143, 14287–14299 (2021).

Article CAS PubMed Google Scholar

Hussain, Z. et al. Repeated recovery of rare earth elements using a highly selective and thermo-responsive genetically encoded polypeptide. Adv. Funct. Mater. 32, 2109158 (2022).

Article MathSciNet CAS Google Scholar

Agafonova, N. V., Kaparullina, E. N., Grouzdev, D. S. & Doronina, N. V. Hansschlegelia quercus sp. nov., a novel methylotrophic bacterium isolated from oak buds. Int. J. Syst. Evol. Microbiol. 70, 4646–4652 (2020).

Article CAS PubMed Google Scholar

Renner, M., Danielson, M. A. & Falke, J. J. Kinetic control of Ca(II) signaling: tuning the ion dissociation rates of EF-hand Ca(II) binding sites. Proc. Natl Acad. Sci. USA 90, 6493–6497 (1993).

Article ADS CAS PubMed PubMed Central Google Scholar

Zallot, R., Oberg, N. & Gerlt, J. A. The EFI web resource for genomic enzymology tools: leveraging protein, genome, and metagenome databases to discover novel enzymes and metabolic pathways. Biochemistry 58, 4169–4182 (2019).

Article CAS PubMed Google Scholar

Lidstrom, M. E. in The Prokaryotes Vol. 2 (eds Dworkin, M. et al.) 618–634 (Springer, 2006).

Mattocks, J. A., Tirsch, J. L. & Cotruvo, J. A. Jr. Determination of affinities of lanthanide-binding proteins using chelator-buffered titrations. Methods Enzymol. 651, 23–61 (2021).

Article CAS PubMed Google Scholar

Pol, A. et al. Rare earth metals are essential for methanotrophic life in volcanic mudpots. Environ. Microbiol. 16, 255–264 (2014).

Article CAS PubMed Google Scholar

Wehrmann, M., Billard, P., Martin-Meriadec, A., Zegeye, A. & Klebensberger, J. Functional role of lanthanides in enzymatic activity and transcriptional regulation of pyrroloquinoline quinone-dependent alcohol dehydrogenases in Pseudomonas putida KT2440. mBio 8, e00570-17 (2017).

Article PubMed PubMed Central Google Scholar

Good, N. M. et al. Hyperaccumulation of gadolinium by Methylorubrum extorquens AM1 reveals impacts of lanthanides on cellular processes beyond methylotrophy. Front. Microbiol. https://doi.org/10.3389/fmicb.2022.820327 (2022).

Grant, T. D. Ab initio electron density determination directly from solution scattering data. Nat. Methods 15, 191–193 (2018).

Article CAS PubMed Google Scholar

Deng, Y. W., Ro, S. Y. & Rosenzweig, A. C. Structure and function of the lanthanide-dependent methanol dehydrogenase XoxF from the methanotroph Methylomicrobium buryatense 5GB1C. J. Biol. Inorg. Chem. 23, 1037–1047 (2018).

Article CAS PubMed PubMed Central Google Scholar

Habenschuss, A. & Spedding, F. H. The coordination (hydration) of rare earth ions in aqueous chloride solutions from x‐ray diffraction. II. LaCl3, PrCl3, and NdCl3. J. Chem. Phys. 70, 3758–3763 (1979).

Article ADS CAS Google Scholar

Hu, A., MacMillan, S. N. & Wilson, J. J. Macrocyclic ligands with an unprecedented size-selectivity pattern for the lanthanide ions. J. Am. Chem. Soc. 142, 13500–13506 (2020).

Article CAS PubMed PubMed Central Google Scholar

Choi, T. S. & Tezcan, F. A. Overcoming universal restrictions on metal selectivity by protein design. Nature 603, 522–527 (2022).

Article ADS CAS PubMed PubMed Central Google Scholar

Altschul, S. F. et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucl. Acids Res. 25, 3389–3402 (1997).

Article CAS PubMed PubMed Central Google Scholar

Teufel, F. et al. SignalP 6.0 predicts all five types of signal peptides using protein language models. Nat. Biotechnol. 40, 1023–1025 (2022).

Article CAS PubMed PubMed Central Google Scholar

Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 13, 2498–2504 (2003).

Article CAS PubMed PubMed Central Google Scholar

R Core Team. R: A Language and Environment for Statistical Computing. http://www.R-project.org/ (R Foundation for Statistical Computing, Vienna 2018).

Edgar, R. C. High-accuracy alignment ensembles enable unbiased assessments of sequence homology and phylogeny. Nat. Commun. https://doi.org/10.1038/s41467-022-34630-w (2022).

Kalyaanamoorthy, S., Minh, B. Q., Wong, T. K. F., von Haeseler, A. & Jermiin, L. S. ModelFinder: fast model selection for accurate phylogenetic estimates. Nat. Methods 14, 587–589 (2017).

Article CAS PubMed PubMed Central Google Scholar

Minh, B. Q. et al. IQ-TREE 2: new models and efficient methods for phylogenetic inference in the genomic era. Mol. Biol. Evol. 37, 1530–1534 (2020).

Article CAS PubMed PubMed Central Google Scholar

Bouckaert, R. et al. BEAST 2.5: an advanced software platform for Bayesian evolutionary analysis. PLoS Comput. Biol. 15, e1006650 (2019).

Article CAS PubMed PubMed Central Google Scholar

Yu, G., Smith, D. K., Zhu, H., Guan, Y. & Lam, T. T. ggtree: an R package for visualization and annotation of phylogenetic trees with their covariates and other associated data. Methods Ecol. Evol. 8, 28–36 (2017).

Article Google Scholar

Featherston, E. R., Mattocks, J. A., Tirsch, J. L. & Cotruvo, J. A. Jr. Heterologous expression, purification, and characterization of proteins in the lanthanome. Methods Enzymol. 650, 119–157 (2021).

Article CAS PubMed Google Scholar

Gasteiger, E. et al. in The Proteomics Protocols Handbook (ed. Walker, J. M.) 571–607 (Humana, 2005).

Manalastas-Cantos, K. et al. ATSAS 3.0: expanded functionality and new tools for small-angle scattering data analysis. J. Appl. Cryst. 54, 343–355 (2021).

Article CAS Google Scholar

Franke, D. et al. ATSAS 2.8: a comprehensive data analysis suite for small-angle scattering from macromolecular solutions. J. Appl. Cryst. 50, 1212–1225 (2017).

Article CAS Google Scholar

Konarev, P. V., Volkov, V. V., Sokolova, A. V., Koch, M. H. J. & Svergun, D. I. PRIMUS - a Windows-PC based system for small-angle scattering data analysis. J. Appl. Cryst. 36, 1277–1282 (2003).

Article CAS Google Scholar

Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326 (1997).

Article CAS PubMed Google Scholar

McCoy, A. J., Storoni, L. C. & Read, R. J. Simple algorithm for a maximum-likelihood SAD function. Acta Crystallogr. D 60, 1220–1228 (2004).

Article PubMed Google Scholar

Zwart, P. H. et al. Automated structure solution with the PHENIX suite. Methods Mol. Biol. 426, 419–435 (2008).

Article CAS PubMed Google Scholar

Grosse-Kunstleve, R. W. & Adams, P. D. Substructure search procedures for macromolecular structures. Acta Crystallogr. D 59, 1966–1973 (2003).

Article CAS PubMed Google Scholar

Terwilliger, T. C. et al. Iterative model building, structure refinement and density modification with the PHENIX AutoBuild wizard. Acta Crystallogr. D 64, 61–69 (2008).

Article CAS PubMed Google Scholar

Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D 60, 2126–2132 (2004).

Article PubMed Google Scholar

Afonine, P. V. et al. Towards automated crystallographic structure refinement with phenix.refine. Acta Crystallogr. D 68, 352–367 (2012).

Article CAS PubMed PubMed Central Google Scholar

Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D 66, 213–221 (2010).

Article CAS PubMed PubMed Central Google Scholar

Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D 66, 12–21 (2010).

Article CAS PubMed Google Scholar

Doutch, J., Hough, M. A., Hasnain, S. S. & Strange, R. W. Challenges of sulfur SAD phasing as a routine method in macromolecular crystallography. J. Synchrotron Rad. 19, 19–29 (2012).

Article CAS Google Scholar

Zheng, H. et al. CheckMyMetal: a macromolecular metal-binding validation tool. Acta Crystallogr. D 73, 223–233 (2017).

Article CAS Google Scholar

Chaudhuri, D., Horrocks, W. D., Amburgey, J. C. & Weber, D. J. Characterization of lanthanide ion binding to the EF-hand protein S100β by luminescence spectroscopy. Biochemistry 36, 9674–9680 (1997).

Article CAS PubMed Google Scholar

Supkowski, R. M. & Horrocks, W. D. On the determination of the number of water molecules, q, coordinated to europium(III) ions in solution from luminescence decay lifetimes. Inorg. Chim. Acta 340, 44–48 (2002).

Article CAS Google Scholar

Seitz, M., Oliver, A. G. & Raymond, K. N. The lanthanide contraction revisited. J. Am. Chem. Soc. 129, 11153–11160 (2007).

Article CAS PubMed PubMed Central Google Scholar

Cramer, R. E., Rimsza, J. M. & Boyle, T. J. The lanthanide contraction is a variable. Inorg. Chem. 61, 6120–6127 (2022).

Article CAS PubMed Google Scholar

Drake, S. K., Lee, K. L. & Falke, J. J. Tuning the equilibrium ion affinity and selectivity of the EF-hand calcium binding motif: substitutions at the gateway position. Biochemistry 35, 6697–6705 (1996).

Article CAS PubMed Google Scholar

Download references

This work was financially supported primarily by Department of Energy (DOE) grant DE-SC0021007 (to J.A.C.), as well as by NSF grant CHE-1945015 (to J.A.C.), supporting the Hans-LanM and Mex-LanM work, respectively. A.K.B. acknowledges National Institutes of Health grant GM119707 and C.-Y.L. acknowledges a fellowship from the Jane Coffin Childs Memorial Fund for Medical Research. D.M.P., C.S.K.-Y. and Z.D. acknowledge financial support from the Critical Materials Institute, an Energy Innovation Hub funded by the DOE, Office of Energy Efficiency and Renewable Energy, Advanced Materials and Manufacturing Technologies Office. Part of this work was carried out under the auspices of the DOE by Lawrence Livermore National Laboratory under contract DEAC52-07NA27344 (LLNL-JRNL-840467). This research used resources of the Advanced Photon Source, a DOE Office of Science User Facility operated by Argonne National Laboratory under contract number DE-AC02-06CH11357. Use of the Life Sciences Collaborative Access Team Sector 21 was supported by the Michigan Economic Development Corporation and the Michigan Technology Tri-Corridor (grant 085P1000817). N.H.Y. acknowledges National Institutes of Health SIG awards S10-OD0215145 for isothermal titration calorimetry instrumentation, S10-OD028589 for the SAXS instrumentation and S10-OD030490 for the SEC–MALS–dynamic light scattering system. We thank J. Fecko for experimental assistance and Y. Jiao and G. Deblonde for discussions.

Department of Chemistry, The Pennsylvania State University, University Park, PA, USA

Joseph A. Mattocks, Jonathan J. Jung, Chi-Yun Lin, Emily R. Featherston, Timothy A. Hamilton, Amie K. Boal & Joseph A. Cotruvo Jr

Critical Materials Institute, Physical and Life Sciences Directorate, Lawrence Livermore National Laboratory, Livermore, CA, USA

Ziye Dong, Christina S. Kang-Yun & Dan M. Park

The Huck Institutes of the Life Sciences, The Pennsylvania State University, University Park, PA, USA

Neela H. Yennawar

Department of Biochemistry and Molecular Biology, The Pennsylvania State University, University Park, PA, USA

Amie K. Boal

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

You can also search for this author in PubMed Google Scholar

J.A.C. identified Hans-LanM and conceived and directed the study. J.A.M. purified Hans-LanM and carried out most biochemical analyses, with assistance from T.A.H. J.J.J. crystallized the proteins, and J.J.J. and C.-Y.L. solved the structures with input from A.K.B. N.H.Y. carried out SAXS studies and supervised some biophysical data collection. E.R.F. purified Mex-LanM. C.S.K.-Y. carried out bioinformatics analyses. Z.D. carried out metal separation experiments, with input from D.M.P. J.A.M., N.H.Y., A.K.B. and J.A.C. wrote the paper with inputs from all authors. All authors edited and approved the final manuscript.

Correspondence to Dan M. Park, Amie K. Boal or Joseph A. Cotruvo Jr.

J.A.M., J.J.J., C.-Y.L., Z.D., E.R.F., C.S.K.-Y., D.M.P., A.K.B. and J.A.C. are inventors on a patent application filed by The Pennsylvania State University based on this work.

Nature thanks Scott Banta and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

The Hans cluster includes LanMs from bacteria from genera Hansschlegelia, Ancylobacter, Methylopila, Oharaeibacter, Starkeya, and Xanthobacter. Although these genera are restricted to this cluster, members at the family level are found dispersed throughout the network, including one Xanthobacteraceae and 42 Methylocystaceae.

Both DyIII (up to 0.3 µM) and CaII (up to 5.5 mM) induce a similar, incomplete conformational change in the protein, relative to the conformational change induced by NdIII and LaIII. The data in the right panel of a is a representative titration from the 3 datasets used to generate the plot in the left panel. The data in b and c are representative titrations from the 3 datasets used to generate the plot in Fig. 1d. Conditions: 15 μM protein, 20 mM acetate, 100 mM KCl, 10 mM EDTA (for Ca and Nd titrations) or EGTA (for Dy titration), 0—10 mM metal ion. Each data point in (a, left panel) is the mean ± s.d. for three independent measurements.

Source Data

a, Overall structure of the asymmetric unit, which consists of two Hans-LanM dimers and two citrate molecules from the crystallization solution. The structure of each monomer of the dimer is consistent with the NMR solution structure of YIII-bound Mex-LanM with EF-hands 2 and 3 paired and EF-hands 1 and 4 paired5. b–e, Details of metal coordination in the four EF-hands of LaIII-Hans-LanM. The coordination spheres of the LaIII ions in EF-hands 1, 2, and 3 are constituted by the side chain Oδ of N1 (monodentate), the carboxylate side chains from D3, D5, E9, and E12 (all bidentate), and a backbone carbonyl from S7 (EF1) or T7 (EF2 and EF3), for a total coordination number of 10. All LaIII-ligand distances are 2.5-2.7 Å. The crystal radius for 10-coordinate LaIII is given as 1.41 Å by Shannon;7 given 1.26 Å as radius of 6-coord O2-, 2.57 Å is estimated for the LaIII-O distance, consistent with our results. The metal ion in EF4 was modeled as NaI because of the shorter metal-ligand distances, lower coordination number, and the presence of sodium in the crystallization solution. CaII cannot be completely ruled out as it was present earlier in the protein purification; however, the protein was treated with Chelex at the end of the purification, and the crystallographic data were consistent with the NaI assignment as determined by the CheckMyMetal server65. This ion is coordinated in distorted pentagonal bipyramidal geometry by monodentate D1, N3, and D5 sidechains, the bidentate E12 sidechain, the backbone carbonyl of Lys113, and a single solvent molecule for a total coordination number of 7. The NaI-protein ligand distances are 2.3-2.5 Å, with a solvent molecule at 2.7 Å. In the case of Mex-LanM, biochemical data4,26 and NMR spectroscopy5 have also supported EF4 as a poor lanthanide-binding site, and it was modeled without a metal ion in the NMR solution state structure5.

a, One of the dimers in the asymmetric unit, comprising chains A and B. Note that EF4 is unexpectedly occupied with DyIII while EF1 is occupied only in chain A. b, Overall structure of the asymmetric unit, which consists of two Hans-LanM dimers. Unlike the LaIII-Hans-LanM structure, the two dimers – and the monomers within each dimer – display significant differences in DyIII-Hans-LanM. EF2-4 are occupied by DyIII in all chains, whereas EF1 is only occupied and ordered in chain A; in chains B and C, no metal ion is bound in the EF-hand, and in chain D, a DyIII ion is bound but the first five residues of EF1 (Asn34 – Asp38) could not be modeled. Our decision to model DyIII into all four EF-hands is supported by anomalous diffraction datasets (Supplementary Tables 9–10, Supplementary Figs. 25–26). The biochemical data suggest that, in solution, at least one DyIII binding site is weak (see Fig. 1d and Supplementary Fig. 2), and it is likely based on studies of Mex-LanM that EF2/3 are the tighter binding sites. This proposal is supported by the Dy anomalous data (Supplementary Table 10), and the occupancy of weak metal-binding sites likely results from the high protein concentration used for crystallography. c, Details of metal coordination in the EF-hands of DyIII-Hans-LanM. In the top row, the three different EF1 structures in the asymmetric unit are shown. Only in chain A is the EF1 metal site nearly identical to the sites in EF2 and EF3 (contrary to LaIII-Hans-LanM, where EF1-3 sites are very similar, Extended Data Fig. 3). In EF1 (chain A), EF2, and EF3, the coordinating ligands are the same as with LaIII-Hans-LanM, except that the E9 residues (Glu42, Glu66, and Glu91) have shifted to monodentate coordination, resulting in 9-coordination. The lower coordination number with DyIII is consistent with the lanthanide contraction68,69 and is observed with other ligands (as one recent example, ref. 39). The DyIII-ligand distances are mostly 2.3-2.5 Å, ~0.2 Å shorter than for LaIII-Hans-LanM. Consistent with this observation, the crystal radius for 9-coordinate DyIII is given as 1.22 Å by Shannon7, 0.19 Å shorter than for 10-coordinate LaIII (Extended Data Fig. 3). The carboxylate shift of the 9th position Glu residue is noteworthy as this position is important for gating affinity and selectivity in other EF-hand proteins70. In EF4, DyIII is 7-coordinate with pentagonal bipyramidal geometry, similar to the sodium site in LaIII-Hans, but with slightly shorter metal-ligand distances (2.2-2.5 Å); again, these distances are consistent with the expectation for 7-coordinate DyIII (ref. 7).

Emission values are normalized to 1.0 for the fluorescence of the apoprotein. Note that the fluorescence intensity of Hans-LanM's Trp residues decreases going from the RE-bound to apo state (Supplementary Fig. 30), whereas the intensity of Mex's Tyr residue increases going from the RE-bound to apo state4,6. Initial conditions: 20 μM protein, 40 μM RE, 20 mM acetate, 100 mM KCl, pH 5.0, for all experiments, into which increasing concentrations of citrate were titrated. The citrate concentrations at which 50% of each metal is desorbed under these conditions ([citrate]1/2) are summarized in Supplementary Table 11 and plotted in Fig. 4a. a, Hans-LanM. b, Hans-LanM(R100K). The compressed difference between the [citrate]1/2 values for La and Nd in Hans-LanM(R100K) vs. wild-type Hans-LanM illustrates the role of dimerization in enhancing affinity differences for the LREs, especially LaIII. c, Mex-LanM. Nd and Dy data were reported in Dong et al6. d, Comparison of the ratios of [citrate]1/2 value for Nd to that for Dy, for each protein, illustrating the greater Nd/Dy selectivity of Hans-LanM relative to Mex-LanM. The ratios for wild-type Hans-LanM and the R100K variant are not significantly different (p > 0.05) by two-tailed t-test, suggesting that the hydrogen-bonding network involving Arg100 contributes relatively little to NdIII/HRE selectivity, though it does impact LaIII selectivity significantly (Fig. 4a). All data are shown as mean ± s.d. (a—c) or s.e.m. (d) for data from 3 independent experiments.

Source Data

The desorption scheme consisted of three stepped concentrations of malonate (30, 50, 90 mM; see right axis) followed by pH 1.5 HCl. The results revealed that slightly lower purity Dy was generated using Hans-LanM compared to the R100K variant (83.6% vs 98% Dy purity at similar yield, respectively; compare to Fig. 4d). While similar selectivity profiles were observed for the immobilized proteins for La through Gd in equilibrium binding experiments with La-Dy, the selectivity pattern diverged at Tb (Fig. 4c). The selectivity difference between Hans-LanM and the R100K variant was confirmed by using a Nd/Dy binary system, as the uncertainties in the distribution factor determination for Dy in the 9-element RE group precluded the ability to distinguish small differences in the Dy/Nd separation factor between proteins (Supplementary Tables S13–S14). In this binary Nd/Dy experiment (Extended Data Table 3), we determined a separation factor of 8.12 ± 0.40 for Hans-LanM and 12.7 ± 1.3 for the R100K variant, which is consistent with the improved Dy separation efficacy of R100K. While consistent with the values derived from the 9-element experiment, the results differ slightly from the equilibrium binding results with the free Hans-LanM and Hans-LanM(R100K) proteins, which revealed similarly high selectivity for Nd over Dy (Fig. 4a,b), likely reflecting weaker LRE-induced dimerization in the R100K variant at the low protein concentration (20 µM) of the solution experiments with free protein. The La/Nd selectivity on-column is also distinct from that observed with the apparent Kd values of the free proteins (wild-type and R100K) in solution, although the experiments with free proteins utilized single element solutions and effects from mixed metal binding may impact the on-column data. The R100K variant is also better behaved on the column, as evidenced by the 2:1 RE:protein stoichiometry. One possible explanation for these results could be that immobilization interferes with dimerization; however, Fig. 2b shows the N- and C-termini of the Hans-LanM dimer, indicating that the C-termini are ~20 Å from the nearest part of the dimer interface, suggesting that immobilization per se would not be expected to disrupt this interface. It must be considered, however, that a functional dimer would require two C-termini to be immobilized in close proximity, which is unlikely at the immobilization densities of our columns. Therefore, on balance, we suspect that the dimerization equilibrium is only applicable in a minority of protein units immobilized on the column. We posit that more fully exploiting the dimerization equilibrium in the column format would yield even more robust separations. The surest way to obtain homogeneous populations of dimers on-column would likely be to link two monomers together (e.g., with a polypeptide chain), tuning dimerization affinity through mutagenesis of the residues contributing to inter-monomer interactions, and immobilizing this dimer through a single attachment point. Dimerization could also be exploited in other separation formats. These directions are the subject of current efforts.

Source Data

This file contains Supplementary Methods, Supplementary Figs 1-31, Supplementary Tables 1-16 and Supplementary References.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions

Mattocks, J.A., Jung, J.J., Lin, CY. et al. Enhanced rare-earth separation with a metal-sensitive lanmodulin dimer. Nature 618, 87–93 (2023). https://doi.org/10.1038/s41586-023-05945-5

Download citation

Received: 29 December 2022

Accepted: 13 March 2023

Published: 31 May 2023

Issue Date: 01 June 2023

DOI: https://doi.org/10.1038/s41586-023-05945-5

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.