A Protein Engineered to Bind Uranyl Selectively and with Femtomolar Afﬁnity

ARTICLES PUBLISHED ONLINE: 26 JANUARY 2014 | DOI: 10.1038/NCHEM.1856

A protein engineered to bind uranyl selectively and with femtomolar afﬁnity

Lu Zhou1†, Mike Bosscher1†, Changsheng Zhang2,3†,SalihO¨ zc¸ubukc¸u1, Liang Zhang1, Wen Zhang1, Charles J. Li1, Jianzhao Liu1, Mark P. Jensen4, Luhua Lai2,3* and Chuan He1*

21 ∼ Uranyl (UO2 ), the predominant aerobic form of uranium, is present in the ocean at a concentration of 3.2 parts per 109 (13.7 nM); however, the successful enrichment of uranyl from this vast resource has been limited by the high concentrations of metal ions of similar size and charge, which makes it difﬁcult to design a binding motif that is selective for uranyl. Here we report the design and rational development of a uranyl-binding protein using a computational screening process in the initial search for potential uranyl-binding sites. The engineered protein is thermally stable and offers very high afﬁnity and selectivity for uranyl with a Kd of 7.4 femtomolar (fM) and >10,000-fold selectivity over other metal ions. We also demonstrated that the uranyl-binding protein can repeatedly sequester 30–60% of the uranyl in synthetic sea water. The chemical strategy employed here may be applied to engineer other selective metal-binding proteins for biotechnology and remediation applications.

ranium is the key element for nuclear-energy production and Through billions of years of evolution nature has produced is important in many other applications. The most stable and strategies to recognize beneficial or toxic metal ions with high Urelevant uranium ion in aerobic environments is the uranyl sensitivity and selectivity. In many cases, with the help of a well- 2þ cation, UO2 . As a consequence of its solubility the concentration folded protein scaffold and the assistance of second-sphere of uranyl in sea water is surprisingly high at 3.2 mg per tonne of sea interactions, metal ions can be recognized in the femtomolar to water (13.7 nM)1. Based on this concentration the ocean is esti- zeptomolar (10221) range with extremely high selectivity14–16.Ifa mated to contain 1,000 times more uranium than land contains, correct and robust scaffold can be developed for the metal of inter- and so offers an enormous resource that, unlike land resources, est, such affinities are sufficient for economic mining of uranium or may be tapped at minimal environmental cost. other elements from sea water or to remediate polluted environ- Functionalized polymers with amidoxime-type ligands have mental sites. Additionally, proteins may be displayed on the surfaces been used to sequester uranyl in sea water since the 1980s2,3. of living organisms17, thus allowing for the biological regeneration However, the low binding affinity and selectivity of the ligands of these systems for recovery and/or remediation purposes at very used limited the utility of the approach. Other approaches based low costs. on functional groups or small-molecule chelates have been pro- Uranyl-binding motifs are well known and offer distinctive posed, with limited success4,5. Sea water is slightly basic and con- handholds for the rational design of proteins that selectively bind tains 2.2 mM total carbonate, which chelates uranyl strongly uranyl7,8. Uranium prefers to oxidize to the þ6 state with two and leaves a free uranium concentration of only 2 × 10217 M axial oxo ligands and form the linear triatomic uranyl ion with (Supplementary Tables 1 and 2, and Supplementary Fig. 1). To an overall charge of þ2. The ability to afford five or six equatorial compete with this high concentration of dissolved carbonate ligands in pentagonal or hexagonal bipyramidal geometry separates under practical conditions, a ligand that can bind uranyl with uranyl from most of the alkali, alkaline and transition metals. The close to femtomolar (fM, 10215) affinity is required (see presence of the axial oxo ligands as potential hydrogen-bond Supplementary Information). In addition, calcium(II) is present at acceptors also distinguishes uranyl from most known lanthanide 10 mM in sea water and prefers an oxygen-rich environment and actinide species in the environment7. Although some uranyl- similar to that which typically binds uranyl. Therefore, a daunting binding motifs have been designed in organic ligands, in DNA selectivity of 106-fold for uranyl over calcium is desired. and in proteins6–11, none have been able to cross the affinity Sophisticated chelating ligands6–11 have been developed for other threshold to compete with carbonate (2.2 mM in ocean) and useful applications, but high synthetic costs limit their utilization achieve the selectivity requirements over other metals in sea when dealing with vast amounts of sea water. A layered solid-state water. Our strategy was to use computational screening and ion exchanger, K2MnSn2S6, has been shown to have high affinity design to develop a stable protein with sufficient binding affinity for uranium in addition to many other metals12. Very recently, for uranyl and selectivity over other metals. The protein can be metal–organic frameworks (MOFs) have also been employed as immobilized on a solid support or displayed on the cell surface novel sorbents to extract uranyl from aqueous media13. Biological for repeated use and sequestration (Fig. 1). This approach can be systems, which can be self-regenerated, offer an opportunity to applied to both engineered systems (resin immobilization) and achieve economically both the required affinity and high selectivity. biological systems (cell-surface display).

1Department of Chemistry and Institute for Biophysical Dynamics, The University of Chicago, 929 East 57th Street, Chicago, Illinois 60637, USA, 2BNLMS, State Key Laboratory for Structural Chemistry of Unstable and Stable Species, College of Chemistry and Molecular Engineering and Center for Quantitative Biology, Peking University, Beijing 100871, China, 3Center for Life Sciences, Peking University, Beijing 100871, China, 4Chemical Sciences and Engineering Division, Argonne National Laboratory, Argonne, Illinois 60439, USA, †These authors contributed equally to this work. *e-mail: [email protected]; [email protected]

236 NATURE CHEMISTRY | VOL 6 | MARCH 2014 | www.nature.com/naturechemistry

NATURE CHEMISTRY DOI: 10.1038/NCHEM.1856 ARTICLES

Pible et al. designed an algorithm to identify native uranyl- 2+ 2+ 19 UO2 VO binding proteins and the corresponding binding sites . However, Zn2+ Ni2+ the Amber force-field-based scoring and optimizing algorithm Ca2+ Fe3+ used in their study was not suitable for searching the vast 2+ 2+ Cu Mg numbers of mutations shown in our de novo design. The scoring function (see Supplementary Information) used in our program Adsorption not only predicts correct uranyl coordination geometry, but also Recovery helps to establish potential hydrogen-bonding interactions between uranyl oxo and residues from the binding protein. Structures of uranyl complexes with inorganic or organic ligands8, as well as Solid support Solid support crystal structures of proteins in which uranyl was used as a source 2+ UO2 of anomalous signal (Supplementary Table 3), provided a basis for the screening and uranyl-binding protein designs. Figure 1 | Uranyl sequestration strategy. Immobilization of a high-affinity Using URANTEIN we searched the PDB for pockets that could and selective uranyl-binding protein on a solid support allows for the accommodate hexagonal bipyramid or pentagonal bipyramid enrichment of uranium over other metals. The protein was de novo designed uranyl-binding geometries, either natively or through mutation of by computational screening based on the known scaffolds in the PDB. potential ligand residues to aspartate/asparagine or glutamate/gluta- Immobilized on resin or displayed on the cell surface by chemical or mine (see Supplementary Figs 2–4 and Supplementary Table 3). We biological means, the protein could effectively sequester uranyl from sea set the optimum distance between the oxygen-based ligand from the water or uranyl-containing groundwater. protein to the central uranium ion at 2.46 Å. In addition, we also incorporated a search method for potential hydrogen bonding Results between the oxo groups of the uranyl with the scaffold protein in Computational screening for potential uranyl-binding sites. We our screening process to maximize potential uranyl-binding affinity developed a large-scale computational screening algorithm, and selectivity (see Supplementary Information). From this initial named URANTEIN, to search the Protein Data Bank (PDB) for screen we identified over 5,000 hits. Hits were further selected pockets that may accommodate uranyl (Fig. 2). A total of 12,173 based on their potential stability, potential steric clashes in the pre- protein structures in the PDB were considered as scaffolds, on dicted coordination site and accessibility of the binding site. The which every residue could be mutated to Asp, Glu, Asn or Gln carboxylate side-chain ligand can be bidentate or monodentate, using the corresponding rotamers in silico. Based on an algorithm and therefore the fivefold or sixfold planar coordination around modified from Automatch18, URANTEIN can search efficiently the central uranyl can be flexible and affected by surrounding for uranyl-binding sites from these mutated pseudo-proteins. protein residues. Ten promising candidates were selected, of

Step 1 Step 2 Step 3

Coordination features Pick a scaffold Build oxygen library and hydrogen library

Step 5 Step 4 Go back to step 2 if there is no hit

Scoring and sorting

Top solutions library Search for uranyl-binding sites

Figure 2 | The main steps in computational screening and design of uranyl-binding proteins. In step 1, three uranyl-coordination features were designed. In step 2, protein scaffolds were prepared by PDB query (version 2010). All entries that contained protein chains of length 60–200 amino acids (12,173 in total) were employed as scaffolds. Then, one scaffold in the library was picked for further screening (for example, the green one shown here). In step 3, the oxygen library of possible mutations and the hydrogen library of native protein were built (Supplementary Fig. 3). In step 4, for each scaffold, uranyl-binding sites were searched using the oxygen library and hydrogen library. An algorithm called URANTEIN was used to complete this work efﬁciently (see Supplementary Fig. 3 for the detailed algorithm). In step 5, the uranyl coordination geometries for the selected sites were evaluated and the results were ﬁltered based on the evaluating score. Finally, all the results were sorted based on the scores.

NATURE CHEMISTRY | VOL 6 | MARCH 2014 | www.nature.com/naturechemistry 237

ARTICLES NATURE CHEMISTRY DOI: 10.1038/NCHEM.1856

a b 108

1.0 107

106 0.8 K d = 7.4 fM 105 0.6 104

3 0.4 10

102 Fraction uranyl bound to protein 0.2 Molar excess in sea water/selectivity 101

0.0 100 10–5 10–4 10–3 10–2 10–1 Na+ Mg2+ K+ Ca2+ Sr2+ Rb+ Ba2+ VO2+ Pb2+ Ni2+ Zn2+ Cu2+ Hg2+ Cd2+ Fe3+ Mn2+ Co2+ Total carbonate concentration (M) Metal ions

Figure 3 | Uranyl-binding affinity and selectivity of SUP. a, Competition assay of SUP versus total carbonate for uranyl yields a Kd of 7.4 fM at pH 8.9. The 2þ final solution of each point contained 10 mMprotein,10mMUO2 and different concentrations of carbonate (fit parameters: maximum ¼ 0.992+0.013, 2 2 minimum ¼ 0.033+0.010, midpoint ¼ 0.398+0.018 mM, Hill coefficient ¼ 1.72+0.11, x red ¼ 0.457, R ¼ 0.9934). b, Binding selectivity of SUP for uranyl over various other metals relevant to sea water extraction. Hatched columns, molar excess of ions in sea water; filled columns, selectivity of metal ions to uranyl by competition assay. The maximum molar excess tested was 4 × 106 for most metals because of the solubility of the salts and limits of detection of the assay. For mercury and lead, the maximum molar excess tested was 105 (for experimental details, see Supplementary Information and Supplementary Table 6). The standard deviations were calculated from the uncertainty in the measurement of uranyl concentration. which nine proteins expressed well and four showed binding to instance, SUP showed selectivity for uranyl greater than or equal × 6 uranyl with Kd values of 100 nM and below (Supplementary to 2 10 -fold over calcium(II). To our knowledge such a high Tables 4 and 5). selectivity is unprecedented for proteins. For those ions that did compete, none competed at concentrations low enough to interfere Development of super uranyl-binding protein (SUP). One of with binding to uranyl in sea water. In fact, the only metals that we the hits stood out for its potential stability and utility, a small could show to compete are Cu2þ at 103-fold excess and vanadyl protein with unknown function from Methanobacterium (VO2þ)at104-fold excess. The concentration of Cu2þ is roughly thermoautotrophicum, an anaerobe isolated from sewage sludge in 2.4 nM in sea water and the concentration of vanadium is about 20 a Urbana, Illinois . This protein consists of three -helices in a 40 nM, and therefore neither could significantly affect the uranyl tight bundle (PDB accession 2PMR, Supplementary Fig. 4) and is binding. The high affinity for Cu2þ may not be surprising given thermally stable at room and elevated temperatures. In the the overall stability of Cu2þ complexes as described in the Irving– computational model, Asp68, Asn17Glu, Leu13Asn and His64Gln William series. The vanadyl ion has similarities in structure and coordinate with the uranium atom, and Arg71 forms a hydrogen charge to the uranyl ion, with only a slightly smaller covalent bond with one uranyl oxo group. Three mutations of Leu13Asn, radius, which suggests a binding motif similar to that of uranyl. Asn17Glu and His64Gln were introduced in the wild-type protein (Supplementary Fig. 4). The mutant protein was cloned, expressed Crystal structure of SUP. To investigate further the mechanism of and purified. It exhibited a modest binding affinity for uranyl uranyl binding by SUP, we crystallized both uranyl-bound SUP and with a Kd of 37 nM (U09, see Supplementary Information). Based the apo-SUP at pH 4.0 (Fig. 4, Supplementary Table 7). The high- on the model structure, we designed mutations that may increase resolution crystal structure of the uranyl–protein complex showed the binding affinity of U09. Although we were able to achieve a a pentagonal bipyramidal binding configuration similar to the modest increase in binding affinity (Kd ¼ 1.8 nM) by mutating a initial computational model (Fig. 4b,d). The uranium position nearby leucine to threonine (Leu67Thr) to stabilize the complex only deviates by 0.47Å (Fig. 4d). Three of the designed ligand structure, we were pleased to find that further mutations of the residues (Glu17, Asp68 and Arg71) interact with uranyl to show two neutral computational mutations His64Gln and Leu13Asn to side-chain conformations similar to the predicted conformations. Glu and Asp, respectively, led to an almost 106-fold increase of Major deviations occur in the modifications of Asn13Asp and × 215 the uranyl-binding Kd to 7.4+2.0 fM at pH 8.9 (7.4 10 M, Gln64Glu designed post-screening, as these two residues are not Fig. 3a, Supplementary Table 5). At pH 6.0, the binding affinity of directly involved in uranyl binding in the solved structure SUP for uranyl decreases to Kd ¼ 0.2+0.1 nM (Supplementary obtained under acidic conditions. We cannot exclude the Fig. 5 and Supplementary Table 5), which implies that the two possibility that these residues may bind to the central uranyl residues may contribute significantly to uranyl binding only at under the basic conditions used for binding assays and uranyl higher pH. The additional negative charge introduced by the sequestration from sea water; however, we have yet to crystallize mutations could also help to stabilize binding of the positively the protein under basic or neutral conditions. Overall, the crystal charged uranyl. structure of the uranyl–protein complex supports the validity of As SUP exhibits a high binding affinity to uranyl, we examined the URANTEIN algorithm as an effective tool for the initial its selectivity for uranyl over that for other metal ions. The screen. Binding of uranyl at the interface of three monomers was protein was immobilized on sulfhydryl resin (see Supplementary also observed (Supplementary Fig. 6). Most probably, this site is Information). Seventeen metal ions found in sea water and relevant formed during the crystallization process because the protein to competition with uranyl were competed against uranyl for forms a stable dimer in solution, whether the uranyl ion is protein binding (Fig. 3b, Supplementary Table 6). Most ions were present or not (Supplementary Fig. 7). In addition, this secondary not able to outcompete uranyl, even at 2.0 × 106-fold excess. For binding site does not appear to be important in the binding of

238 NATURE CHEMISTRY | VOL 6 | MARCH 2014 | www.nature.com/naturechemistry

NATURE CHEMISTRY DOI: 10.1038/NCHEM.1856 ARTICLES

a b c d

Arg71 Helix I Arg71 2.9 Asn13 Glu17 Asp13 Glu17 Asp68 Helix III Lys9 Asp68 2.7 2.4 w124 3.1 2.9 2.3 Gln64 Thr67 Helix II Glu64

Crystal structure Design model

Figure 4 | Uranyl–SUP crystal structure. The high-resolution crystal structure of the uranyl–protein complex reveals the metal binding site that is close to the computational prediction. The uranyl ion is in a negatively charged pocket with a pentagonal bipyramidal binding configuration. a, Overview of the crystal structure of the SUP–uranyl complex. b, Uranyl-binding pocket detail of SUP. c, SUP surface electrostatics (blue colour for the positive regions and red colour for the negative regions). d, Comparison of SUP design model (magenta) to SUP crystal structure (blue). uranyl, as titration of 10 mM SUP with uranyl shows the presence of fusion protein bound to amylose resin showed a good efficiency, a single strong binding site (Supplementary Fig. 5). with 17% of the uranyl sequestered with one equivalent of Structural features explain the high uranyl-binding affinity and immobilized protein used against one equivalent of uranyl selectivity of SUP, which contains four mutations (Asp13, Glu17, (13 nM) in the corresponding amount of sea water in a very short Glu64 and Thr67), compared to the original template protein: incubation time of 30 minutes (Fig. 5b), whereas ten equivalents of the immobilized protein sequestered 30% of the uranyl from † Residues Glu17 and Asp68 directly bind the uranyl ion in the the same amount of synthetic sea water. When an excess of SUP equatorial plane. Mutations of Glu17Gln or Asp68Asn dramati- proteins (.6,000 equiv.) was used we were able to remove over cally decreased the binding affinity, which supports these resi- 90% of the uranyl (Supplementary Fig. 10), similar to recent dues as being ligands to uranyl. The remaining equatorial results with a layered solid-state ion-exchanger approach12, and binding site is occupied by a water molecule. only less than 5% of the uranyl remained bound to protein-free † Asp13 and Glu64 provide a negatively charged environment to amylose resin. We also used the Escherichia coli surface display as hold the uranyl ion (Fig. 4c). The negative charge is a seemingly a complimentary example of sea water sequestration. SUP protein essential attribute because the neutral mutant homologues was displayed on the E. coli surface using an established OmpA (Asn13, Gln64) show much lower affinity. These residues could fusion method17. The protein was highly expressed and displayed bind uranyl directly under basic conditions. on the cell surface (see Supplementary Fig. 11). This system is † Arg71 forms a hydrogen bond with an axial oxo of the uranyl capable of extracting more than 60% of the uranyl in synthetic sea ion in the complex structure, but in the apo form it forms a salt water (Fig. 5b). The engineered protein is stable thermally with a bridge with Glu17 (Fig. 4b, Supplementary Fig. 8). This inter- Tm of 71 8C (see Supplementary Fig. 12) and the bacterial action contributes to the high selectivity for the uranyl ion system can be regenerated easily, which thus offers an economic through recognition of the oxo group. Vanadyl(IV), like source for uranyl sequestration. To our knowledge, this is the first uranyl, is an oxo cation with a þ2 charge and can compete biological system that can effectively sequester uranyl from sea water. with uranyl at the 104-fold level, which suggests that Arg71 may also form a hydrogen bond with the single oxo group Discussion of vanadyl. Owing to active human activities, natural metal resources have been † Leu67Thr stabilizes the binding state via hydrogen bonding with exploited at an accelerated pace with increasing metal pollution. backbone Ala63. Thr67 also better accommodates the bending of Thus, new technologies that can sequester and remediate metals effi- the last helix on uranyl binding compared to that of leucine ciently from dilute environmental sources and sites are highly desir- because threonine is a residue known to destabilize the helix able. Uranyl, the predominant aerobic form of uranium, is present structure (Supplementary Fig. 9). in the ocean at a concentration of 13.7 nM. To sequester uranyl effectively from sea water requires the development of uranyl- Uranyl sequestration from sea water. To confirm the function of binding ligands with femtomolar or higher affinity for uranyl and 6 this uranyl-binding protein, we tested uranyl-sequestration exceedingly high selectivity over calcium(II)(.1 × 10 -fold) and efficiency. More than 95% of the uranyl was removed after other abundant metal ions present in the ocean. Currently, no exist- treating uranyl-containing water (666 nM uranyl, 158 parts per ing small-molecule or protein-based ligands can achieve this 109 (ppb) at pH 7.0) with SUP fused with a maltose-binding daunting task. protein (MBP-SUP) immobilized on amylose resin (see We present here a strategy to design and develop selective metal- Supplementary Information). The residual uranyl concentration binding proteins starting from a computational screening for suit- was 3–8 ppb, far lower than the 30 ppb US Environmental able scaffolds. Applications of de novo protein design22–24 as well Protection Agency standard for drinking water in the USA. The as rational design based on existing protein scaffolds have led to resin can be recycled easily by washing with carbonate solution metalloproteins with new functions25–31. A rapidly expanding and the efficiency was retained after many cycles (Fig. 5a). knowledge of protein structure and function has allowed researchers Next, we applied MBP-SUP to test uranyl sequestration from to employ effective strategies that can incorporate active sites synthetic sea water. Synthetic sea water was prepared by a ranging from catalytic sites to binding hot spots and metal-coordi- method, reported previously, to account for all major ions and nation patterns27 into new scaffolds for designing novel anions above 1 mM, with uranyl present at 13 nM (ref. 21). The enzymes32,33, target protein binders34,35 and metalloproteins36,37.

NATURE CHEMISTRY | VOL 6 | MARCH 2014 | www.nature.com/naturechemistry 239

ARTICLES NATURE CHEMISTRY DOI: 10.1038/NCHEM.1856

a 100 b 100

80 80

60 60

40 40

20 20 Per cent uranyl sequestered (%) Per cent sea water recovered (%)

0 0 1 2 3 4 5 6 7 8 9 Fusion protein Fusion protein Cell surface to uranyl 1:1 to uranyl 10:1 display Cycle

Figure 5 | Immobilized SUP provides a useful manifold for a variety of applications. a, SUP immobilized on amylose resin can consistently remove uranyl from uranyl-contaminated water over many cycles. These experiments were performed in pH 7.0 buffer to mimic groundwater. The standard deviations were calculated from the uncertainty in the measurement of uranyl concentration. b, Both protein display by fusion with maltose-binding protein and immobilization on amylose resin and by fusion with OmpA and display on live E. coli cells allow for effective recovery of uranyl in synthetic sea water21.The use of one and ten equivalents of resin-immobilized protein against one equivalent of uranyl (13 nM) in the synthetic sea water can sequester 17% and 30% of the total uranyl, respectively. Over 60% of the uranyl can be sequestered with the use of surface-displayed E. coli cells. The standard deviations were calculated from triplicate experiments.

Central to this strategy are the determination of active-site structures conditions required for potential applications. The constructs and the scaffold selection. Although the incorporation of metal- we describe open up broad possibilities for the extraction of binding sites to both de novo and native scaffolds has been uranyl and for biological remediation. The strategy presented studied comprehensively36, nearly all of these metalloprotein- here may be applied to engineer protein-based reagents for design cases utilized either known scaffolds from existing or de selective metal binding in metal sequestration and remediation. novo designed metalloprotein templates. Although large-scale scaffold selection has been employed successfully in enzyme and Methods protein–protein interaction designs32–37, such a selection strategy Computational design of uranyl-binding proteins. All entries from the PDB that has yet to be applied for metalloprotein designs. contain protein chains of length 60–200 amino acids (12,173 in total) were We have shown that a de novo design program can be developed employed as scaffolds. Uranyl coordination was generated based on uranyl- containing protein structures from the PDB and uranyl-binding molecules from the to search for specific sites in scaffold proteins that may accommo- Cambridge Crystallographic Data Centre (CCDC). The oxygen ligands in the date the pentagonal or hexagonal bipyramidal geometry that equatorial plane of the hexagonal bipyramidal or pentagonal bipyramidal uranyl binds uranyl selectively from most other metal ions. The subsequent could come from main-chain amides, water, natively or mutationally generated rational design based on structural and binding experiments aspartate/asparagine or glutamate/glutamine; mutations were generated using the (Supplementary Fig. 2) led to a stable protein that exhibits femto- basic Mayer rotamer library. A program named URANTEIN was developed to search the protein scaffold library efficiently for pockets that might accommodate molar affinity and unprecedented selectivity towards uranyl uranyl. The program obtained all possible coordinating oxygens (which generated binding. This robust system, which can be regenerated on bacterial an oxygen library), and then searched for uranyl-binding sites using the library (see surfaces or reused on the resin, is capable of practically enriching details in Supplementary Information and Supplementary Fig. 3). The uranyl- uranyl from sea water and both avoids competition from major binding sites were evaluated by a scoring function that contained oxygen ions and repeatedly cleansing uranyl-containing water. Although coordination, oxygen compatibility and hydrogen bonding. The best-scoring solutions were checked further manually with consideration of steric clash, the depth other approaches to extract uranyl from sea water have been pro- of uranyl in its binding pocket and the stability of the scaffold protein. posed, and even implemented, our protein-based system may address three limitations present in current strategies: (1) cost, (2) Protein purification and crystallization. The gene sequences of U01-U10 and SUP- selectivity and (3) adaptability: OmpA fusion protein were synthesized by GeneScript. Mutations were performed using Pfu Ultra II polymerase from Agilent. U09 and mutations were cloned and expressed in pMCSG19 vector for expression in BL21 (DE3) pRK1037 E. coli strains. (1) The genetically encoded system described here may be biologi- The strains that carried the plasmids were grown in LB (Lysogeny Broth) to cally regenerated and displayed on the surface of living bacteria attenuance (D600) ¼ 0.6, induced with 1 mM IPTG, and cells were grown overnight or in plants, with minimal costs. The bacteria or plants may be at room temperature before harvesting. Cells were lysed by sonication in the lysis buffer that contained 20 mM Tris-HCl (pH 7.4), 500 mM NaCl and 1 mM used directly for uranyl enrichment. dithiothreitol (DTT) in the presence of 1 mM phenylmethyl sulfonyl fluoride as the (2) The superb selectivity is characteristic of protein-based binding. serine protease inhibitor. Supernatant was separated by centrifugation and loaded on Such a high selectivity would be cost prohibitive to achieve with nickel nitrilotriacetic acid columns. Protein was obtained using 10 mM Tris pH 7.4, large-scale synthetic systems based on chelator or ion exchange. 500 mM NaCl with 1 mM DTT as elution buffer and imidazole ramping from 0 to Once developed, the superb selectivity and the high binding affi- 500 mM. The His6-tag was removed by tobacco etch virus (TEV) protease. All crystallization samples were further purified by gel filtration in buffer containing nity of the current system are genetically encoded and readily 10 mM Tris pH 7.4, 100 mM NaCl and 1 mM DTT. Crystals were produced using regenerated in living host systems. Other approaches may not hanging-drop vapour diffusion at 16 8C by mixing 1 ml protein solution at 10–20 mg match the selectivity of our protein-based system in the presence ml21 with a 1 ml reservoir solution that contained 2% v/v Tacsimate pH 4.0, 0.1 M of many other metal ions in sea water. sodium acetate trihydrate pH 4.6, 16% polyethylene glycol 3350. For the SUP–uranyl (3) Furthermore, with a genetically encoded system, directed evol- complex, uranyl was mixed with SUP at a 1.2:1 molar ratio before the drop set. ution could be employed in the future to improve the perform- Uranyl-binding assay. A modification of the Arsenazo III method was employed to ance further and to make the system adaptable to various determine uranyl concentrations. Arsenazo III) (80 mM, 50 ml) that contained 0.1 M

240 NATURE CHEMISTRY | VOL 6 | MARCH 2014 | www.nature.com/naturechemistry

NATURE CHEMISTRY DOI: 10.1038/NCHEM.1856 ARTICLES

HCl was titrated with an equal volume of uranyl solutions ranging from 0 to 30 mM, 18. Zhang, C. & Lai, L. Automatch: target-binding protein design and enzyme and the absorbances at 652 nm and 800 nm were monitored. The values of A652 and design by automatic pinpointing potential active sites in available protein A800 increased linearly in this range, and can be converted into uranyl scaffolds. Proteins 80, 1078–1094 (2012). concentrations. Diglycolic acid (DGA) and carbonate competition assays were used 19. Pible, O., Guilbaud, P., Pellequer, J. L., Vidaud, C. & Que´meńeur, E. Biochimie to determine the Kd values of uranyl-binding protein. DGA competition assays were 88, 1631–1638 (2006). performed at pH 6.0 or 6.5 in 10 mM Bis-Tris buffer with 300 mM NaCl. Standard 20. Zeikus, J. G. & Wolee, R. S. Methanobacterium thermoautotrophicus sp. n., an 2þ solutions of 100 mM protein and 100 mMUO2 were prepared and diluted tenfold anaerobic, autotrophic, extreme thermophile. J. Bacteriol. 109, 707–715 (1972). with the appropriately scaled DGA buffer. The final solutions contained 10 mM Bis- 21. Saito, K. & Miyauchi, T. Chemical forms of uranium in artificial seawater. 2þ Tris (pH 6.5), 10 mM protein, 10 mMUO2 and different concentrations of DGA. J. Nucl. Sci. Technol. 19, 145–150 (1982). The solutions were mixed and filtered through centrifuge filters with a 3 kDa cutoff. 22. Regan, L. & DeGrado, W. F. Characterization of a helical protein designed from Flow through was tested for uranyl concentration using an Arsenazo III assay. first principles. Science 241, 976–978 (1988). Carbonate competition assays were performed similarly to the DGA assays, but with 23. Lovejoy, B. et al. Crystal structure of a synthetic triple-stranded alpha-helical freshly prepared pH 8.0–9.0 carbonate solutions in Tris-HCl buffer. All the water bundle. Science 259, 1288–1293 (1993). used to prepare solutions was freshly degassed, deionized and protected against 24. Wendt, H., Berger, C., Baici, A., Thomas, R. M. & Bosshard, H. R. Kinetics of further sorption of atmospheric CO2. The Kd of the uranyl–protein complex was folding of leucine zipper domains. Biochemistry 34, 4097–4107 (1995). determined by fitting the resulting binding curve using the uranyl–DGA or uranyl– 25. Pordea, A. & Ward, T. R Artificial metalloenzymes: combining the best features carbonate binding constants. of homogenous and enzymatic catalysis. Synlett 20, 3225–3236 (2009). 26. DeGrado, W. F., Summa, C. M., Pavone, V., Nastri, F. & Lombardi, A. De novo Sea water sequestration assays by amylose resin or surface-display cells. For the design and structural characterization of proteins and metalloproteins. Annu. preparation of SUP-bound amylose resin, SUP was cloned into pMCSG19 vector Rev. Biochem. 68, 779–819 (1999). and expressed in BL21 E. coli to produce a MBP fusion protein. The protein was then 27. Dutton, P. L. & Moser, C. C. Engineering enzymes. Faraday Discuss. 148, incubated with amylose resin to obtain SUP-bound resin. For the preparation of 443–448 (2011). surface-displayed cells, the SUP-OmpA fusion protein was cloned into pBAD vector 28. Khare, S. D. et al. Computational redesign of a mononuclear zinc metalloenzyme and expressed in BL21 E. coli, with a linker GGGSGGGS between SUP and OmpA. for organophosphate hydrolysis. Nature Chem. Biol. 8, 294–300 (2012). Cells were grown to D ¼ 0.6 in LB medium and induced by 2% L-arabinose 29. Azoitei, M. L. et al. Computation-guided backbone grafting of a discontinuous overnight, and harvested by centrifugation at 5,500g. For the sea water extraction motif onto a protein scaffold. Science 334, 373–376 (2011). assay, SUP-bound amylose resin (5 ml) or surface-display cells (0.5 ml pellet) were 30. Radford, R. J., Brodin, J. D., Salgado, E. N. & Tezcan, A. Expanding the utility of incubated with 50 ml synthesized sea water for 30 minutes at room temperature. proteins as platforms for coordination chemistry. Coord. Chem. Rev. 225, Supernatants were sent for uranium analysis by inductively coupled plasma 790–803 (2011). mass spectrometry. 31. Brodin, J. D. et al. Metal-directed, chemically tunable assembly of one-, two- and three-dimensional crystalline protein arrays. Nature Chem. 4, 375–382 (2012). Received 7 November 2013; accepted 17 December 2013; 32. Saven, J. G. Computational protein design: engineering molecular diversity, published online 26 January 2014 nonnatural enzymes, nonbiological cofactor complexes, and membrane proteins. Curr. Opin. Chem. Biol. 15, 452–457 (2011). References 33. Privett, H. K. et al. Iterative approach to computational enzyme design. Proc. 1. Nozaki, Y. A fresh look at element distribution in the North Pacific. Eos Trans. Natl Acad. Sci. USA 109, 3790–3795 (2012). Am. Geophys. Union 78, 221–227 (1997). 34. Liu, S. et al. Nonnatural protein–protein interaction-pair design by key residues 2. Tabushi, I., Kobuke, Y. & Nishiya, T. Extraction of uranium from seawater by grafting. Proc. Natl Acad. Sci. USA 104, 5330–5335 (2007). polymer-bound macrocyclic hexaketone. Nature 280, 665–666 (1979). 35. Fleishman, S. J. et al. Computational design of proteins targeting the conserved 3. Kanno, M. Present status of study on extraction of uranium from sea water. stem region of influenza hemagglutinin. Science 332, 816–821 (2011). J. Nucl. Sci. Technol. 21, 1–9 (1984). 36. Lu, Y., Yeung, N., Sieracki, N. & Marshall, N. M. Design of functional 4. Egawa, H., Nonaka, T. & Nakayama, M. Recovery of uranium from seawater. 7. metalloproteins. Nature 460, 855–862 (2009). Concentration and separation of uranium in acidic eluate. Ind. Eng. Chem. Res. 37. Touw, D. S., Nordman, C. E., Stuckey, J. A. & Pecoraro, V. L. Identifying 29, 2273–2277 (1990). important structural characteristics of arsenic resistance proteins by using 5. Sather, A. C., Berryman, O. B., Ajami, D. & Rebek, J. Jr Selective recognition and designed three stranded coils. Proc. Natl Acad. Sci. USA 104, extraction of the uranyl ion. J. Am. Chem. Soc. 132, 13572–13574 (2010). 11969–11974 (2007). 6. Franczyk, T. S., Czerwinski, K. R. & Raymond, K. N. Stereognostic coordination chemistry. 1. The design and synthesis of chelators for the uranyl ion. J. Am. Acknowledgements Chem. Soc. 114, 8138–8146 (1992). This work was supported by the Division of Chemical Sciences, Geosciences, and 7. Gordon, A. E., Xu, J., Raymond, K. N. & Durbin, P. Rational design of Biosciences, Office of Basic Energy Sciences of the US Department of Energy, under sequestering agents for plutonium and other actinides. Chem. Rev. 103, contract number DE-FG02-07ER15865 to C.H., and at Argonne National Laboratory 4207–4282 (2003). (M.J.) under contract number DE-AC02-06CH11357, the Dreyfus Foundation 8. Gong, H. Y., Rambo, B. M., Karnas, E., Lynch, V. M. & Sessler, J. L. A Postdoctoral Program in Environmental Chemistry to S.O., the Ministry of Science and ‘Texas-sized’ molecular box that forms an anion-induced supramolecular Technology of China (2009CB918500) and the National Natural Science Foundation of necklace. Nature Chem. 2, 406–409 (2010). China (21173013, 11021463) to L.L. Use of the Advanced Photon Source for protein 9. Wegner, S. V., Boyaci, H., Chen, H., Jensen, M. P. & He, C. Engineering a crystallography data collection at beamlines LS/CA-CAT (21-ID-F) and NE-CAT uranyl-specific binding protein from NikR. Angew. Chem. Int. Ed. 48, (24-ID-C) was supported by the Office of Basic Energy Sciences of the US Department of 2339–2341 (2009). Energy under contract number DE-AC02-06CH11357. We thank S. F. Reichard for 10. Lee, J. H., Wang, Z., Liu, J. & Lu, Y. Highly sensitive and selective colorimetric editing the manuscript and C. Yang and L. Lan for experimental support. 2þ sensors for uranyl (UO2 ): development and comparison of labeled and label-free DNAzyme–gold nanoparticle systems. J. Am. Chem. Soc. 130, Author contributions 14217–14226 (2008). C.H. conceived the project. C.H. and L.L. designed the experiment. C.Z. and L.L. performed 11. LeClainche, L. & Vita, C. Selective binding of uranyl cation by a novel the initial screening. W.Z. and C.J.L. performed subcloning of virtual hits. S.O. and C.J.L. calmodulin peptide. Environ. Chem. Lett. 4, 45–49 (2006). expressed, purified and tested all first-generation virtual hits. L.Z., S.O. and M.B. designed 12. Manos M. J. & Kanatzidis M. G. Layered metal sulfides capture uranium from second-generation mutants. L.Z., M.B. and J.L. expressed and tested all protein derivatives seawater. J. Am. Chem. Soc., 134 (39), 16441–16446 (2012). and designed all later-generation mutants. L.Z., M.B., M.P.J. and C.H. analysed the data and 13. Carboni, M., Abney, C. W., Liu, S. & Lin, W. Highly porous and stable metal– M.B. and L.Z. co-wrote the manuscript. All authors discussed the results and commented organic frameworks for uranium extraction. Chem. Sci. 4, 2396–2402 (2013). on the manuscript. 14. Changlea, A. et al. Molecular basis of metal-ion selectivity and zeptomolar sensitivity by CueR. Science 301, 1383–1387 (2003). 15. Aisen, P., Leibman, A. & Zweier, J. Stoichiometric and site characteristics of the Additional information binding of iron to human transferrin. J. Biol. Chem. 253, 1930–1937 (1978). Supplementary information is available in the online version of the paper. Reprints and 16. Banci, L. et al. Affinity gradients drive copper to cellular destinations. Nature permissions information is available online at www.nature.com/reprints. Correspondence and 465, 367–368 (2010). requests for materials should be addressed to L.L. and C.H. 17. Georgiou, G. et al. Display of beta-lactamase on the Escherichia coli surface: ′ ′ outer membrane phenotypes conferred by Lpp –OmpA –beta-lactamase Competing financial interests fusions. Protein Eng. 9, 239–247 (1996). A patent application has been filed for the technology disclosed in this publication.

NATURE CHEMISTRY | VOL 6 | MARCH 2014 | www.nature.com/naturechemistry 241

© 2014 Macmillan Publishers Limited. All rights reserved. Copyright of Nature Chemistry is the property of Nature Publishing Group and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.