The rare earth elements (REEs) -a family of elements comprising the 15 lanthanides, plus yttrium and scandium -possess similar physiochemical properties and play indispensable roles in the emerging green economy. 1 With increasing technological dependence on these elements, however, the chemical, environmental, and political challenges associated with mining and processing REEs have been magnified. [2][3][4] These complexities have driven interest in obtaining REEs more sustainably from low-grade but abundant non-traditional sources, such as coal byproducts, mine effluents [e.g., acid mine drainage (AMD)], and recycling from electronic waste (E-waste). [5][6] The discovery that biology is able to selectively recognize and utilize the lighter lanthanides, especially La-Nd, [7][8][9][10][11][12] offers new possibilities for efficient biotechnologies to meet these challenges. 13 This promise has been accelerated with the recent identification of the first native, selective biological chelator for the lanthanides, lanmodulin (LanM). 14 This small (12-kDa) protein undergoes a conformational change that is 10 8 -fold more selective for lanthanides (with a preference for the lighter REEs) over non-REEs tested to date. [14][15] The protein tolerates acidic conditions (pH<2) relevant to environmental REE streams and industrial processes, and it is able to quantitatively extract REEs from acidic coal and E-waste leachates with high purity, outperforming traditional chelators. 15 From a chemical perspective, this selectivity for REEs is all the more remarkable because the protein utilizes EF hands, carboxylate-rich metal-binding motifs associated with Ca II recognition in most of the hundreds of other characterized examples. [16][17] An understanding of the fundamentals underlying LanM's selective recognition of REEs may translate not only to more efficient approaches for REE extraction and separations but also to simpler and portable methods of REE detection (also necessary for ensuring a sustainable supply of REEs), as we demonstrate in the present work.
Biochemical characterization of Methylorubrum extorquens LanM 14 suggested that binding of REEs to two of the protein's three metal-binding sites occurs cooperatively with a picomolar apparent Kd, inducing formation of two-thirds of the protein's helical content; a third site binds with similar but slightly weaker affinity, contributing the remaining one-third. The NMR structure of lanmodulin, obtained in the presence of the diamagnetic Y III ion, revealed the overall structure of the protein, the three REE-binding EF hands (EF1-3), and a hydrophobic core that may help stabilize the REE-bound state of the protein (Figure 1), but it left many questions unanswered. 18 Although the protein architecture is unusual for an EF-hand protein, the pairing of EF2 and EF3 suggested that they might account for the cooperative metal-binding phase.
Alternatively, the unusual connection of EF1 and EF2 through a single, short α-helix made stabilization of this helix via cooperative metal binding to EF1 and EF2 also a plausible source of the cooperative phase (Figure S1). Furthermore, the detailed structure of the sites -in particular, whether solvent molecules contributed to the coordination spheres of the metal ions -could not be determined from the NMR structure. Therefore, the means by which structural and kinetic properties of the metal sites contribute to LanM's unique REE affinity and selectivity still remain to be determined. In order to answer these questions, it is necessary to probe each metal-binding site specifically. Several lanthanide(III) ions (Ln III ) display intrinsic luminescence with relatively long luminescence lifetimes, especially in the cases of Tb III and Eu III , and diagnostic emission spectra. 19 Lanthanide f-f transitions are Laporte forbidden and therefore direct excitation is inefficient; this limitation can be overcome by incorporating a photosensitizer adjacent to the metal ion to absorb and transfer energy to the metal excited state (luminescence resonance energy transfer, LRET). 20- 24 The requirement for a nearby sensitizer is advantageous in that it enables probing of individual metal-binding sites utilizing chromophores (e.g., tyrosine or tryptophan) in the protein, either native or incorporated via site-directed mutagenesis. [25][26] This strategy has been employed extensively in proteins for characterization of metal binding affinity, 27 deduction of binding order and metal-metal distances in multi-metallic systems, [28][29][30] dissociation kinetics, [31][32][33] and solvent coordination, 34 especially in Ca II -binding EF-hand proteins, due to similarities in ionic radius and coordination preferences of Ca II and Ln III ions, such as Tb III . However, this method has yet to be applied to understand a natural, dedicated lanthanide-binding biomolecule.
In addition to their utility in biology and as biochemical probes, sensitive methods for detection of REEs are essential both in the identification of potential waste streams for REE recovery and in monitoring of industrial REE processing operations. Inductively coupled plasma mass spectrometry (ICP-MS) is the gold standard but expensive and not portable, whereas portable instruments like x-ray fluorescence spectrometers are limited by interferences and low sensitivities. [35][36] As recently comprehensively reviewed, 37 luminescence-based methods are attractive alternatives, especially in the case of Tb III , which is both readily sensitized and one of the rarest and most valuable REEs. 38 Terbium is one of five REEs deemed by the United States as the most economically critical, for its uses including in phosphors in energy-efficient lighting, 39 as well as for its domestic supply risk. 40 Despite recent progress to decrease limits of detection (LODs) for Tb III even at low pH and in AMD matrix, [41][42] popular approaches utilizing small molecules or metal-organic frameworks (MOFs) still lack the necessary selectivity, requiring spiking of samples with Tb III for detection. By contrast, biomolecular luminescence-based sensors for Tb III based on Trp-containing EF hands such as lanthanide-binding tags (LBTs) exhibit better selectivity but their affinities (Kd ~50 nM [43][44] or higher 45 at pH 7) are insufficient for applications below pH ~6. 46 As a step toward more sensitive detection of REEs, our laboratory recently reported a selective FRET-based sensor for REEs (LaMP1), taking advantage of LanM's large conformational response. 47 LaMP1 facilitated discovery of key elements of lanthanide uptake machinery in bacteria, but its use is limited to near-neutral pH values, it cannot distinguish between REEs, and it exhibits a small but significant response to some non-REEs at high concentrations.
Consequently, it is unsuitable for complex environmental samples. We hypothesized that the combination of LanM's high affinity and selectivity for lanthanides with sensitized luminescence might allow for efficient and specific detection of terbium. However, low-grade natural sources of REEs like AMD are acidic and contain very low REE concentrations, <1 ppm, and Tb III at ppb levels (1 ppb Tb III ≈ 6 nM). 6 Indeed, to the best of our knowledge, no luminescence-based sensors have been successfully deployed to quantify, or even detect, terbium at natural levels in these matrices.
Here, we show how strategic insertion of tryptophan residues into lanmodulin enables definition of key aspects of the protein's selective REE recognition, including metal site-specific thermodynamics, kinetics, and structure. These insights also suggest how the protein might be optimized further for biotechnological applications. Furthermore, we show that one of these Trp-LanM variants enables detection and quantification of Tb III levels directly in AMD, a challenging matrix inaccessible to previously characterized luminescent sensors. Together, our work suggests that this technology could be extended for detection of other luminescent f-elements and that LanM might enable harvesting of REEs from AMD.
General considerations. Terbium(III) chloride hexahydrate (99.9%) and general laboratory chemicals for protein expression and purification and buffer preparation were obtained from Millipore Sigma. Deuterium oxide (99.9%) was purchased from Cambridge Isotope Laboratories.
Primers were ordered from Integrated DNA Technologies (IDT). E. coli strains [5alpha, BL21(DE3)] for cloning and recombinant protein expression, respectively, as well as cloning reagents (Q5 DNA polymerase, OneTaq Quick-Load, KLD Enzyme Mix, DpnI) were purchased from New England Biolabs. Miniprep kits were from Omega Bio-tek. Protein gel electrophoresis was carried out using Invitrogen Novex WedgeWell 16% Tris-Glycine gels and a mini gel apparatus. Chelex 100 resin was purchased from BioRad. Automated protein chromatography was carried out on a GE Healthcare Biosciences Äkta Pure fast protein liquid chromatography (FPLC) system. UV-visible absorption spectra were obtained on an Agilent Cary 60 UV-visible spectrophotometer using a quartz cuvette (Starna Cells). Well plates were analyzed using a BioTek Synergy H1 microplate reader. Fluorescence titrations and lifetime determinations were carried out on Cary Eclipse and PerkinElmer FL6500 spectrofluorometers, respectively, using a quartz septum cell micro fluorometer cuvette (10 mm pathlength, Starna Cells). Circular dichroism measurements were carried out in the X-ray Crystallography and Automated Biological Calorimetry Facility at Penn State using a 1-mm pathlength quartz CD cuvette (Jasco J/0556).
Stopped flow UV-vis measurements were made on an Applied Photophysics SX20 spectrophotometer, equipped with a 450 long-pass filter (Corion LL-450-F-T539) and a fluorescence detector. All protein and metal solutions were made in 2 mL microcentrifuge tubes or 15 mL or 50 mL centrifuge tubes purchased from Sarstedt. All thermodynamic and kinetic data were analyzed, and curve fitting was performed, in Origin 2018.
The AMD sample was collected from the feed of an AMD treatment facility operated by the Pennsylvania Department of Environmental Protection (Pennsylvania, USA). The source -the same as that used recently by Pisupati and coworkers 6 , albeit collected at a different time -was from the lower Kittanning coal seam. The metal content of the sample was analyzed using inductively coupled plasma mass spectrometry (ICP-MS) on a Thermo Fisher Scientific ICAP RQ (ICP-MS) at the Penn State College of Earth and Mineral Sciences, Earth and Environmental Systems Institute, Laboratory for Isotopes and Metals in the Environment. The AMD sample was diluted 20× into 2% HNO3 (Aristar Ultra, BDH VWR Analytical), and a blank of 2% HNO3 was subtracted from each analyte prior to elemental content determination. The pH of the sample was 3.24.
Mutagenesis of T41, T65, N87, T90, K94, and T114 to W was performed on pET24a-LanM using the forward and reverse primers shown in Table S1, using NEB KLD enzyme mix. Reactions were run for 4 h at room temperature and then supplemented with 2 U/µL DpnI for 30 min at 37 °C prior to transformation. Transformants were screened for insert by colony PCR (OneTaq Quick-Load) and the correct insert was confirmed by DNA sequencing by Genewiz, yielding the plasmids in Table S2. The variants were expressed and purified as described for the wt protein, and stored in 20 mM MOPS, 100 mM KCl, 5 mM acetate, 5% glycerol, pH 7.0 (Buffer A). 48 Determination of optimal Trp insertion points for LRET. Initial screening for the most suitable positions for Trp placement within LanM's EF hands utilized the EF3 Trp-substituted variants, N87W, T90W, and K94W. Stoichiometric LRET titrations of 20 µM protein in 20 mM MOPS, 100 mM KCl, 5 mM acetate, pH 7.0 (Buffer B) were carried out with a 2 mM solution of TbCl3 in the same buffer. Titrations were performed on a Cary Eclipse fluorescence spectrophotometer using a 10 mm pathlength quartz septum cell micro fluorometer cuvette (Starna Cells) with the following instrument parameters: 280 nm excitation, 400-700 nm emission scan, 5 nm excitation and emission slit widths, 120 nm/min scan rate, 1 nm data interval, 250-395 nm excitation filter, 430-1100 nm emission filter, and high PMT voltage setting. A blank solution of buffer was subtracted from each spectrum prior to analysis and spectra were corrected for volume change prior to plotting. Substituting the seventh position in EF3 (T90W) revealed the greatest magnitude of LRET signal at 545 nm (Figure S2), and this position was selected as the optimal position for Trp substitutions in each EF hand (T41W, T65W, T90W, T114W), from the perspective of signal intensity.
Circular dichroism spectroscopy on wt-LanM and Trp-LanM variants. Circular dichroism (CD) spectra of wt LanM and Trp-LanM variants were collected using a Jasco J-1500 CD spectrometer, thermostatted at 25 °C, using a 1-mm pathlength quartz CD cuvette. Samples were scanned from 260-190 nm, with the following instrument settings: 1.00 nm bandwidth, 0.5 nm data pitch, 50 nm/min scan rate, 4 s average time. The cuvette contained 15 µM protein in 200 µL Chelex-treated 30 mM MOPS, 100 mM KCl, pH 7.2 (Buffer C), into which 1 to 5 equivalents TbCl3 were titrated, and spectra were acquired. Three scans were acquired and averaged for each condition. A buffer blank spectrum was subtracted from each sample spectrum, and the spectra were corrected for volume change before plotting.
Kd determinations using CD spectroscopy. Solutions of Tb III with free metal concentration buffered using ethylenediamine N,Nꞌ-disuccinic acid (EDDS), a chelator well matched with LanM's lanthanide affinity, were prepared as described, in Buffer C. 14 Protein was added to the low and high Tb III -EDDS solutions separately to a final concentration of 15 µM, and EDDS solutions were mixed at various high:low ratios. The same ratio of high:low solutions were prepared without protein to yield the blank samples. Following a 1-h incubation at room temperature, 14 samples were scanned from 260-210 nm, with the following instrument settings:
1.00 nm bandwidth, 0.5 nm data pitch, 50 nm/min scan rate, 4 s average time, 25 ºC. One accumulation was acquired for each condition. The blank spectrum acquired for each high:low ratio was subtracted from the corresponding Tb III -LanM spectrum, and [θ]222nm was plotted vs. free metal concentration.
solutions of Tb III were prepared as described above. Protein was added to the low and high Tb III -EDDS solutions separately to a final concentration of 10 µM, and EDDS solutions were mixed at various high:low ratios. Following a 1-h incubation at room temperature, time-resolved fluorescence emission was monitored from 400-700 nm on a BioTek Synergy H1 microplate reader in Greiner Cellstar 96-well half-area µClear plates with the following instrument settings: time-resolved delay 50 µs, collection time 1000 µs, fixed excitation at 295 nm, emission 400-700 nm with 1 nm steps, gain of 120, and read speed delay 200 ms. Data points were corrected for the significant contribution from emission of the Tb III Luminescence lifetimes for q determination. Protein was diluted to 10 µM with 30 µM TbCl3 in Buffer C. For D2O containing samples, protein solutions (3 mL) were lyophilized overnight and resuspended in an equal volume of D2O, the process was repeated, and then 10 µM protein in H2O and 10 µM protein in D2O were mixed in different proportions to achieve the % D2O concentrations desired (0-75%). Lifetime measurements were obtained on a PerkinElmer FL 6500 Fluorometer with the following parameters: data mode phosphorescence (short), excitation correction off, source mode pulse, flash count 1, flash power 120 kW, frequency 50 Hz, excitation wavelength 295 nm, excitation slit 5 nm, excitation filter air, emission wavelength 545 nm, emission slit 5 nm, emission filter air, PMT voltage 700 V, PMT gain auto, emission correction off, response time 0.5 s, delay time 0 µs, gate time 20 ms. Three independent trials were performed for each condition, and the 1/τ values determined from the fitted curves were averaged and plotted against % H2O to determine 1/τ(D2O) from the y-intercept of the line. The values of q were calculated using the method of Horrocks (eq. 1), 34,49 where A = 5.0 ms, and a correction factor of -0.06 ms -1 was applied. 50 (eq. 1)
Stopped-flow fluorometry. Stopped-flow fluorometry measurements were carried out at 25 ºC, maintained by a circulating water bath. One syringe contained a solution of 10 µM Trp-LanM and 30 µM TbCl3, prepared in Chelex-treated Buffer C. Three equivalents of Tb III were used to minimize occupancy of EF4, adjacent to T41W; this stoichiometry is sufficient to fully (T41W) and nearly fully (T90W) saturate the proteins' Tb III -dependent conformational changes (see Results). The contents of this syringe were mixed in a 1:1 ratio with solutions of EGTA (10, 5, 2.5, and 1.25 mM, in Chelex-treated Buffer C) in a second syringe. Data were acquired with the
following parameters: 1 mm slit width, 2 mm pathlength, with excitation at 295 nm, collecting 2000 data points over 120 s (T41W) or 200 s (T90W), and 12.5 µs sample period. Three shots were collected and averaged for each condition. Although a 450 nm long-pass filter was used, there was some residual Trp fluorescence in addition to the LRET signal in the emission channel.
Curve fitting was performed in Origin 2018, fitted to either a single exponential (T41W) or a double exponential (T90W) decay. concentration that displayed a peak across the entire pH range (2.4 or 15.9 ppb, for T90W and T41W, respectively), according to eq. 2.
Quantification of Tb III in AMD using T90W-LanM. Time-resolved luminescence emission was monitored from 400-650 nm in Greiner BioOne 96-well white flat-bottom Lumitrac plates with the same instrument settings as above. The sample volume was 200 µL. With these settings, there was no significant emission from the AMD in the blank sample (without protein). For terbium detection, T90W-LanM was added to a concentration of 10 µM, from a 1 mM stock in Buffer A.
Addition of the protein did not significantly affect pH of the AMD. Tb III concentration was determined by averaging the emission at 544-546 nm. Full luminescence intensity was reached within the time between sample mixing and the first reading (~15 min). To enable Tb quantification, a standard curve was generated by mixing the same volumes of AMD and protein as above but with 0.4-2 µL of 2.5 ppm Tb III in Buffer D added, to yield 5-25 ppb Tb III . The data with 0, 5, 10, 15, 20, and 25 ppb Tb III added were fitted to a regression line, and the emission of the sample without Tb III added was divided by the slope of the line, yielding the estimated Tb III concentration.
Determination of optimal Trp insertion points in LanM. LanM possesses no Trp residues natively, facilitating our strategy to site-specifically probe metal binding using sensitized luminescence. On the basis of the NMR solution structure of Y III -bound LanM, 18 we selected three positions in EF3 -N87 (4 th position), T90 (7 th position), and K94 (11 th position) -for Trp substitution and preliminary assays of energy transfer efficiency. We hypothesized that these substitutions would minimally interfere with metal ion binding yet also be sufficiently close to the Tb III ion to yield a robust LRET signal. These variants were screened by stoichiometric Tb III titrations (pH 7.2), exciting the Trp at 280 nm and monitoring in-growth of the luminescence spectrum of Tb III , in particular the most intense feature at ~545 nm, corresponding to the 5 D4 → 7 F4 transition (Figure S2). In the case of N87W and K94W, Trp emission increased with addition of Tb III , indicative of a change in environment of the fluorophore, presumably associated with the metal-induced conformational change. However, LRET was weak in these variants. Meanwhile, T90W showed the expected quenching of the Trp emission accompanied by strong Tb III emission.
Slight differences in Tb III -binding stoichiometry were observed with these variants. Three equiv.
Tb III was sufficient to maximize Trp response and LRET signal for K94W but not other variants (4 equiv.); by comparison, wild-type LanM binds 3 equiv. REEs with high affinity. 14 Therefore, K94W may introduce the least perturbation into EF3, but it gives a poor LRET signal. As a result, these preliminary studies suggested that the 7 th position Thr residue was the most promising position for Trp substitution from the perspective of LRET efficiency. This result is consistent with characterization of other Trp-substituted EF-hand proteins including calmodulin 29 and synthetic Tb III -binding peptides, 51 including LBTs. 44 Therefore, the 7 th positions of the other three EF hands were substituted with Trp (T41W, T65W, T114W). All of these constructs exhibited robust sensitized Tb III emission (Figure S3), whereas significant sensitized Eu III emission was not observed, similar to other protein systems 49 (Figure S4). Presumably, the response of T114W (EF4) reflects energy transfer to the Tb III bound in EF1.
The Trp-substituted LanM variants (Trp-LanMs) were evaluated by CD spectroscopy to determine whether the Trp residue affected the apparent dissociation constant (Kd,app) and magnitude of the Tb III -induced conformational change. All variants exhibited the same overall conformational change as the wild-type LanM (~2.5-3-fold increase in the molar ellipticity at 222 nm, indicating increased helicity in the presence of Tb III ions), with the notable exception of the EF2 insertion, T65W, which displays less helicity in the apoprotein as well as a nearly completely disrupted conformational response (Figure S5). As a result of the substantial perturbation introduced by the Trp residue in EF2, this site was not directly probed further, although T90W in EF3 does appear to report on metal binding to EF2 as well (vide infra). Disruptive effects resulting from substitution of Trp residues at the EF-hand 7 th position has been observed previously for some Ca II -binding EF-hand proteins. 52 The other variants displayed full conformational response to 3 equiv. Tb III , with the exception of N87W and T90W, which required four equivalents, similar to the fluorescence result (Figure S2). Kd,app values were determined for the Tb III -bound Trp variants in comparison to wt LanM (Figure 2, Table 1). The substitutions most distal to metal-binding sites (K94W in EF3 and T114W in EF4, which does not bind a metal under these conditions) displayed Kd,app values and Hill coefficients (n) very similar to wt LanM. T41W (EF1) displayed a slight increase in Kd,app and decrease in n. By contrast, T90W was disruptive, appearing to break the conformational response into two phases with Kd,app values ~10 pM and ~100 pM; however, two-phase fits did not converge well, so the single-phase fit is reported here. Therefore, T41W and T114W are both suitable probes of metal binding to EF1, and K94W is the least disruptive probe for EF3, although its very weak LRET intensity could limit some experiments. The observation that T65W and (to a lesser extent) T90W, which are both at the interface of EF2 and EF3 (Figure 1), are the most disruptive substitutions also provides important insights into LanM function. First, the failure of T65W-LanM to adopt the full helical structure of the wt protein suggests that EF2 is particularly critical for LanM's conformational change. Second, communication between EF2 and EF3 appears to be important for maintaining high Tb III affinity overall, as suggested by characterization of T90W.
Apparent Kd values and Hill coefficients (n) for Trp-LanMs, determined using CD spectroscopy (Figure 2) and luminescence methods (Figure S6), and monitored at various EDDSbuffered free Tb III 10 ± 1 2.1 ± 0.4 15 ± 1 1.5 ± 0.2 a Note that the Kd,app value for untagged wt LanM determined here is slightly lower (7.3 pM) than for the C-terminally His-tagged protein initially characterized (21 pM); we have shown that the Cterminal tag slightly increases Kd,app values. 14 All studies herein use untagged proteins. b NA: not applicable c T114W appears to report on metal binding to EF1 although its position is within EF4 (see Figure 1D for residue positions within the EF hands).
Having characterized the overall conformational responses of the Trp-LanM variants, we next exploited the long-lived luminescence of Tb III (Figure 3A) using time-resolved detection to probe the individual binding site(s) in the vicinity of each Trp residue. We reasoned that comparison of each of the extracted Kd,app values and cooperativities (Hill coefficient, n) to the CD-derived values for the whole protein could enable deduction of the order of metal binding and connectivity between the three EF hand binding sites. Overall, T41W showed the largest LRET response, followed by T90W and T114W, with K94W the smallest (Figure S6), although all four constructs studied had sufficient responses to allow determination of apparent Kd values and Hill coefficients. T41W and T114W, which both report on Tb III binding to EF1 (Figure 1), exhibited lower Hill coefficients and slightly higher Kd,app values as determined by the LRET titrations, compared to the CD titrations (Table 1).
Because the CD titration of T114W is least perturbed from wt values, this variant likely serves as the better reporter of EF1 binding; therefore, the Kd,app of EF1 is likely ~15 pM (the value derived from LRET), slightly weaker than the main response. These observations suggest that metal binding to EF1 does not account for the main conformational response of LanM (i.e., it is the weakly cooperative second phase). However, because the Hill coefficients associated with the probes of EF1 binding are not exactly equal to 1, it is possible that metal binding to EF1 is slightly influenced to the main binding event (vide infra), perhaps via the helical connectivity between EF1 and EF2 (Figure S1).
Consistent with the CD titrations, T90W is clearly a disruptive position for Trp placement within EF3, as revealed by a Hill cofficient, n≈1, indicating a lack of cooperativity, and a high Kd as compared to wt. Interestingly, the Kd,app value derived from the LRET data approximately corresponds to the apparent second phase in the CD titrations, with Kd,app ~100 pM. These large perturbations meant that we could not draw conclusions about cooperativity of metal binding using this variant. Therefore, the K94W variant, which displayed fully wt behavior in CD titrations, was instead used to investigate Tb III binding to EF3. Although the LRET intensities observed with this variant were very low (Figure 3B, Figure S2C), the apparent Kd was identical to the values determined by CD for this variant as well as for the wt, suggesting that EF3 is involved in LanM's primary metal-induced conformational change. Despite the fact that this Trp residue is predicted to be much closer to the Tb III bound in EF3 (~9 Å) than to any other EF hand (e.g., ~18 Å from EF2), and therefore likely only reports on EF3, the Hill coefficient was ~2, indicating positive cooperativity between Tb III binding to EF3 and to at least one other EF hand. Because EF1 is ruled out from the above considerations, EF3 must be communicating with EF2, supporting the conclusion from the previous section.
Together, these data support the model for metal binding and conformational change presented in Figure 3C. Metal binding events in EF2 and EF3 display strong positive cooperativity and together are responsible for the major conformational change in the protein, accounting for ~2/3 of its helical content. Metal binding to EF1 is slightly weaker and only weakly connected to binding at the other sites, and it is responsible for the remaining 1/3 of the protein's helical content.
Luminescence lifetimes to investigate coordinated waters. Another critical aspect of understanding LanM's function is the structure of the metal binding sites. Our determination of the NMR structure of Y III -bound LanM was unable to determine whether protein residues saturated the metal-binding sites, or whether solvent molecules filled part of the coordination spheres. This information can be obtained from the lifetime of the Tb III excited state, which is sensitive to the presence of coordinated water molecules due to the radiationless decay of the excited state via O-H vibrations. 26,34 Because this decay pathway is suppressed in D2O, the empirical relationship between the difference in decay rate constants (τ -1 ) in the presence of H2O and D2O has been shown to yield the approximate number of coordinated water molecules (q). 34,50 Therefore, we probed LanM using this method, utilizing the T41W, T90W, and K94W variants because of their representation of both pairs of EF hands. In cases, the decays of the luminescence signals could be fitted to single exponentials (Figure 4 for T41W, Figure S7 for T90W, and Figure S8 for K94W). By varying the mole fraction of D2O in the protein solution, the q values determined are 2.0 ± 0.1 for T41W, 1.4 ± 0.1 for T90W, and 2.6 ± 0.1 for K94W. Because of the perturbations introduced by the T90W substitution, we suggest that the K94W data are more likely to represent solvent coordination for the native EF3. It is also possible that the T90W data reflect contributions from both EF2 and EF3. Nevertheless, the data suggest that both EF1 and EF3 metal-binding sites coordinate roughly two solvent molecules. The presence of water molecules coordinated to Tb III or Eu III ions used to probe metal binding in Ca II -binding EF hands proteins is common; 26 for example, luminescence studies have found q = 2 for Ln III -bound metal sites in calmodulin 53 and q = 1 in parvalbumin. 54 Therefore, despite the protein's unique affinity and selectivity for lanthanides, lanmodulin's metal sites are rather typical among EF-hand proteins with regard to solvent coordination. This result raises the question of whether the additional, conserved carboxylate residues at position 9 of LanM's EF hands are in fact coordinated, as originally proposed, or perhaps involved in hydrogen bonding with a coordinated water molecule, as residues at this position sometimes are in Ca II -binding EF hands. 16 Additionally, because the presence of coordinated solvent would be expected to increase the rate constant for metal ion dissociation, we next sought to use stopped flow spectrofluorometry to probe the kinetics of the system in order to better understand this selectivity. By contrast, fitting the T90W decays to a single exponential did not yield acceptable residuals (Figure S9); fitting to two exponential phases was necessary, resulting in koff values of 0.049 ± 0.005 s -1 and 0.020 ± 0.002 s -1 (Table S3). The requirement for two phases to fit the T90W data may reflect the position of this Trp residue between EF3 and EF2, allowing communication to each EF hand; the failure to distinguish a second phase in the luminescence decay experiments above may reflect either the lower signal to noise in the decay experiment or an identical q value for both metal-binding sites. In support of the ability of Trp residues at the 7 th position to communicate with both EF hands in the pair, the T114W (EF4) variant exhibits LRET when Tb III is bound in EF1 (Figure S6, Table 1). Therefore, we suggest that the two phases report on metal dissociation from EF2 and EF3, respectively, but we are unable to assign each phase to a particular EF hand. We speculate that the faster koff may be associated with EF3, based on the influence of T90W on apparent Kds measured by CD and steady-state LRET (Table 1). Of course, such perturbations from introduction of Trp residues may mean that the true koff values for the wt protein may be somewhat smaller than those determined here. Nevertheless, the data suggest that the three metal-binding EF hands in LanM have relatively similar dissociation rate constants, on the order of 0.02-0.05 s -1 .
Even though LanM exhibits several orders of magnitude higher affinities for Ln III ions than most other EF-hand proteins, LanM's koff values are within the typical range for Tb III dissociation from the proteins in this family (e.g., 0.05 and 0.5 s -1 for parvalbumin 31 and 0.01 s -1 for galactose binding protein 32 ). The similarity of these values may be accounted for by the similar numbers of coordinated solvent molecules, as determined above. From the koff values and Kd,app values (Table 1), both determined based on LRET, kon was estimated to be ~8 × 10 8 M -1 s -1 for T41W and 3 -7 × 10 8 M -1 s -1 for T90W (in the case of T90W, a single Kd,app value from the LRET titrations but two different koff values prevents calculation of a single kon). These rate constants are very close to the diffusion limit, which is on the order of 10 9 -10 10 M -1 s -1 . 55 Such rapid association kinetics may help to account for LanM's remarkable ability to selectively bind lanthanides even in very complex solutions, with hundreds of millions-fold excess of competing metal ions. 15 In future work, it will be intriguing to probe kon and koff values for non-REEs. Therefore, the optimization of LanM's kon values for lanthanide binding appears to be a key feature of the protein's metal recognition, whereas its dissociation rates are rather typical among EF-hand proteins. However, control of koff may still be important; the observation that kon is near-optimal even for Tb III , whereas LanM has even higher affinity for the early Ln III ions, suggests that koff differences may govern LanM's selectivity within the lanthanide series, which will be the subject of future work. Experimental curves were fitted to a single (T41W) or double (T90W) exponential decay and rate constants were plotted (mean ± SD, n = 3) were plotted against EGTA concentration. The koff values were determined from the y-intercepts of the linear fits.
to characterize mechanism, structure, and kinetics of LanM was carried out at pH 7.2, potential broader application of these proteins as sensors would require responsiveness under a range of conditions, particularly in the presence of other metal contaminants and at low pH. Although LanM can selectively and quantitatively extract REEs from low-grade feedstocks containing only 30 ppm (~200 µM) total REEs, 15 environmental samples such as AMD typically harbor much lower concentrations, <1 ppm REEs, and Tb III at only low ppb levels. 6 Meanwhile, both these applications and monitoring of industrial processes necessitate robust performance at lower pHs than our previous LaMP1 sensor 47 could provide. With an eye toward these applications, we first determined the pH dependence and limits of detection (LODs) of Trp-LanM luminescence.
We again focused on the T41W and T90W variants, as they exhibited the greatest sensitized luminescence intensity of the Trp-LanM constructs characterized. Standard curves were generated with Tb III concentrations ranging 0.8-36 ppb and between pH 2 and pH 7, in the presence of 1 or 10 µM of each protein. Given the long luminescence lifetime observed (Figure 5), samples were detected in time-resolved luminescence mode, on a plate reader. At pH 3-7, linear responses were observed for both proteins (Figure 6). Neither protein responded at pH 2. Because REEs have been shown to desorb from the protein at pH ~2.5, this result reinforces that the observed luminescence signal is specific to interaction with LanM. Whereas at 1 µM, T90W-LanM exhibits a lower slope at pH 3 than at pH 4-5 (Figure 6B), increasing the concentration to 10 µM results in constant slopes at pH 3 and 4 (Figure 6C), suggesting that the metal is essentially fully bound to protein under these conditions. Interestingly, although T41W performed better than T90W at pH 7, its luminescence declined more quickly at lower pH values than that of T90W. Because wt LanM retains binding of 3 equivalents of REEs even down to pH 3, 15 this decline is unlikely to result from metal dissociation. Instead, we note that the magnitude of the conformational response measured by CD decreases slightly at lower pH values. 15 Together, even though the Kd,app of EF1 is similar to that of EF2/3 at pH 7.2 as demonstrated by our LRET studies (Table 1), these results suggest that EF1 becomes more conformationally labile at lower pH, perhaps because it is paired with EF4, which does not bind metal ions tightly. 2). Even at pH 3, our LOD values are similar to (or better than) the lowest values reported for other complexes, which are often reported at pH 7. 37,42 These results suggested that T90W-LanM might be robust enough to detect Tb III even in challenging environmental samples. Application to quantification of terbium in acid mine drainage. In order to stringently challenge the affinity and selectivity of Trp-LanM and assess its potential for environmental monitoring, we tested the performance of the most promising construct, T90W-LanM, in acid mine drainage. AMD is effluent from active and abandoned mines, by which natural processes leads to release of metals, including REEs, from ores. AMD is often enriched in the rarer and more valuable heavy REEs due to the mechanism of natural acid-based extraction processes. 6 The presence of existing infrastructure required to treat AMD sites and mitigate their environmental impact prior to release of the AMD into natural waters has motivated investigation into feasibility of extracting REEs from these sources, in order to convert waste into revenue streams. [56][57][58] An estimated 770 to 3400 tonnes of REEs are potentially accessible per year from AMD in Pennsylvania and West Virginia (USA) alone; 57 for comparison, current domestic consumption of REEs is estimated at 13,000 tonnes per year. 38 A field-deployable sensor for specific REEs such as terbium, or even a simpler lab procedure for analysis, could enable more rapid assessments of the value of new sites for development for REE extraction, as well as inexpensive in-line monitoring of extraction processes once implemented, compared to current ICP-MS or x-ray fluorescence elemental quantification methods.
We used a samples of AMD (pH 3.24) collected from the feed of an AMD treatment facility in Pennsylvania, USA. 6 Analysis of the sample using ICP-MS showed the presence of 300 ppm total metal ions, including high levels of potential interferents: 239 ppm (9.6 mM) Mg, 25 ppm (0.45 mM) Mn, 19 ppm (0.70 mM) Al, 12 ppm (0.41 mM) Si, and 1.5 ppm (22 µM) Zn (see Table S4 for the full analysis). The sample also contained 280 ppb (2.3 µM) total REEs and only 3.3 ppb (21 nM) Tb III . A sensor concentration of 1 µM did not yield a luminescence signal clearly above background. Although LanM's slight preference for the more abundant early lanthanides may contribute partially to this result, the LODs for T90W-LanM at pH 3 (Table 2) suggest that the primary limitation under these conditions is that the pH of the AMD (3.2) is close to the apparent pKa of the metal sites (~2.5-3) 15 and therefore the Kd is on the same order as the protein concentration. Accordingly, and consistent with Table 2, when 10 µM T90W-LanM -a concentration far below the total metal concentration -was added, the Tb III luminescence features at ~490 and ~545 nm were clearly visible above the background sample with no protein added (Figure 7A). Whereas the combination of the 100-fold excess of REEs over terbium, 100,000-fold excess of non-REEs, and low pH would present a formidable challenge to most luminescencebased Tb III detection methods, T90W-LanM's ability to detect Tb III is consistent with LanM's high affinity for REEs and negligible binding of non-REEs, previously observed. [14][15] Because components of the AMD matrix (other metal ions, as well as anions) could affect quantification, a standard curve was generated in the AMD itself by adding 0-25 ppb Tb III in the presence of 10 µM T90W-LanM (Figure 7B). Because the luminescence of the AMD in the absence of protein is negligible, dividing the intensity of the AMD sample without added Tb (33 ± 11) by the slope of the regression line (8.3) yields a Tb III concentration of 4.0 ± 1.3 ppb (Figure 7C). This value, close to but above the limit of quantification at pH 3, is in excellent agreement with the value determined by ICP-MS, 3.3 ppb. Because the slope of the line is only 50% lower than those of the standard curves obtained in idealized Tb III solutions at pH 3 and pH 4, we suggest that the LODs determined may be limited more so by the sensitivity of the plate reader than by affinity or selectivity of the protein. Using more sensitive detectors (e.g., a spectrofluorometer), lower LODs could conceivably be achieved. Therefore T90W-LanM is an exceptionally sensitive sensor for Tb III , applicable even in complex, environmental samples such as AMD. The performance of Trp-LanM (and T90W-LanM specifically) in detecting Tb III at low concentration even in complex media compares favorably with other luminescence-based sensors, both biomolecular and synthetic. 13,37 Perhaps most conceptually similar to Trp-LanM is the LBT, but its affinity for Tb III is only 60 nM at pH 7 44 (or 3 nM for one site in the "double LBT" construct 59 ). Because this affinity is 3-4 orders of magnitude lower than that of LanM, LBTs would not be expected to function at pH 3. Indeed, attempts to use LBTs for Tb III binding and sensing at pH values below ~5-6 have been unsuccessful. 46 A cell-based sensor incorporating an LBT into a bacterial two-component system responds to as little as ~0.2 µM (30 ppb) Tb III at neutral pH, but also responds significantly to other metals (e.g., Ca II at 50 µM) at concentrations that would be present in environmental samples (e.g., our AMD sample contains 3.11 ppm, or 78 µM, Ca). 60 Numerous synthetic luminescence-based sensors for lanthanides, including terbium, have been characterized, with a wide range of detection limits. 37 Some of these sensors have been characterized in natural samples, although at higher pH values than AMD and spiked with Tb III . 61- 62 A particularly extensively explored approach in recent years utilizes metal-organic frameworks (MOFs), exhibiting better tolerance to lower pHs than LBTs and other sensors but also quenching from interactions with other ions due to low selectivity. 37,63 One of the most promising examples is a zinc-adeninate MOF for detection of several lanthanides (BioMOF-100), yielding an LOD for Tb III of 90 ppb in neutral water. 41 Very recently, Crawford and co-workers have characterized even more sensitive MOFs with LODs of 6 ppb Tb III in neutral water. However, when this sensor was applied in an AMD sample containing ~1 ppb Tb III at pH 3.4, the sample had to be spiked with 800 ppb (5 µM) Tb III to be able to observe a signal. 42 By contrast, simply adding T90W-LanM directly to AMD detects and quantifies 3 ppb Tb III present in AMD at pH 3.24, without any spiking.
This comparison shows that the high affinity and selectivity of LanM for lanthanides provides a significant advantage over conventional REE sensitizer ligands, which (while they may have high affinity), usually do not have sufficient selectivity to work well in complex solutions like AMD.
Indeed, to the best of our knowledge, our work represents the first time that Tb III can be quantified at such low levels in AMD or a similar, low-pH environmental sample using a luminescence-based sensor.
Incorporation of Trp residues site-specifically into LanM's metal-binding sites not only provides insights into the protein's metal recognition but also yields a technology capable of specific detection of terbium, even in complex samples. This work establishes several key aspects of LanM function. First, our results point to EF2 and EF3 as being LanM's preferred metal-binding sites and responsible for the cooperative binding phase. EF1 is slightly lower affinity, but interestingly appears to be more destabilized at low pH values, even though metal binding is retained. 13,15 This result suggests that EF1 might be the most dispensable site for applications of LanM to REE extraction and separations. In addition, we suggest that LanM-bound Tb III ions possess two coordinated solvent molecules. The presence of these coordinated solvent molecules may explain our observation that LanM's association kinetics are extremely and uncommonly fast, perhaps minimizing the energetic penalty for dehydration upon metal binding. [64][65] Rapid kinetics may be useful for separations applications, whereas these results also suggest that it may be possible to mutate the protein to decrease metal dissociation rates for other applications.
Furthermore, our work establishes that EF3, and particularly the T90W position, is an especially fruitful location for installation of sensitizers within the protein scaffold with acceptable perturbation of metal binding and strong responsiveness across a wide pH range. Of the variants tested here, T90W-LanM is optimal for terbium detection under a range of conditions; however, it is easy to imagine a pathway for further optimization and extension of the system. Other lanthanides can be detected besides terbium via their unique luminescence signatures, including other valuable REEs such as Eu III , Dy III , Sm III , and Nd III , 21,37 but Trp is not a suitable sensitizer for these elements, nor is Trp ideal due to its ultraviolet excitation. Our results lay the groundwork for site-specific incorporation of other sensitizers into LanM that would allow sensitive detection of other lanthanides, as well as even more sensitive detection of Tb III . [66][67] In addition, because wt LanM favors early lanthanides, one can envision extension of this approach to LanM variants with distinct selectivity profiles that favor middle-late lanthanides (e.g., perhaps 4P→4A) 14 to further optimize detection. Increasing overall lanthanide affinity of the protein would also be important for enabling metal quantification at the lowest pH values using lower protein concentration (Table 2). Finally, the related coordination chemistry of the trivalent actinides suggests that Trp-LanMs may also serve as efficient sensitizers for some of these elements, such as Cm III . 68 The robust performance of the Trp-LanM sensors builds on our prior work to detect total REEs at neutral pH 47 and to extract REEs under harsh conditions, 15 extending LanM's capabilities to facile detection of a specific REE in real environmental samples, with potential extension to detection of other elements. Furthermore, the detection of terbium in a sample would indicate the likely presence of higher concentrations of other generally more abundant but also valuable rare earths, such as neodymium and dysprosium. The small quantities of protein required, and the adaptability of this system to available luminescence detectors, motivates further exploration and