Restriction enzymes, ligase, and other molecular reagents for gene cloning were obtained from NEB (Ipswich, MA, United States). The pET-24a-ELP[V150] plasmid was purchased from Addgene (ID: 67015) (Watertown, MA, United States). Macrogen, Inc. (Seoul, South Korea) synthesized all DNAs and primers used for cloning. Unless otherwise specified, chemical reagents were obtained from Sigma-Aldrich (Saint Louis, MO, United States).

An FA sample was collected from a coal-fired power plant operated by the Korea Midland Power Boryeong Power Generation Site Division. A measure of 5 g of FA was leached using 50 mL of 2 M HCl solution for 3 h. The leachate was centrifuged at 5,000 g for 15 min to remove any undissolved particles before pH adjustment using NaOH. The pH-adjusted leachate was centrifuged again to remove precipitates formed during pH adjustment. The final leachate solution obtained had a pH of 4.5. The concentrations of Y³⁺, La³⁺, Ce³⁺, Pr³⁺, Nd³⁺, Sm³⁺, Eu³⁺, Gd³⁺, Tb³⁺, Ho³⁺, Er³⁺, Tm³⁺, and Yb³⁺ were determined by ICP-MS (Agilent 7900, Agilent Technologies, Santa Clara, CA, United States). The Na²⁺, Mg²⁺, and Ca²⁺ concentrations were determined by ICP-OES (iCAP7400DUO, Thermo Scientific, Waltham, MA, United States). Samples were diluted with 2% HNO₃ for analysis. The ICP-MS and ICP-OES results corresponded to the mean values (n = 3).

The expression and purification of RELPs were performed as previously reported (Hussain et al., 2022). In brief, the pET-24a-RELP plasmid encoding the RELP gene was constructed by adding the LanM gene to the pET-24a-ELP[V150] plasmid. The features of the pET-24a-RELP plasmid are shown in Supplementary Material. The DNA and amino acid sequences of RELPs were previously reported (Hussain et al., 2022). After the protein expression and purification of RELP, the concentration of the purified RELP (molar extinction coefficient: 6,990 m⁻¹ cm⁻¹) was measured at an absorbance of 280 nm using a microplate reader (Synergy, BioTek, Winooski, VT, United States) according to Beer–Lambert’s law (Pace et al., 1995; Anthis and Clore, 2013).

The AlphaFold Colab notebook from DeepMind was used for protein structure prediction (Jumper et al., 2021), and the full RELP sequence was provided as an input. The structure was visualized using PyMOL Molecular Graphics System, version 1.2r3pre, Schrödinger, LLC.

To investigate the phase transition of the RELP at different concentrations, the RELP in 20 mM MES buffer (pH 5.8) was heated from 26°C to 44°C at a rate of 1°C/2 min, and changes in absorbance at 350 nm (A _350nm ) were monitored using the microplate reader. The reversible phase-transition behavior over multiple cooling and heating cycles was further investigated by measuring size changes using dynamic light scattering (DLS). DLS was performed using the Zetasizer Nano ZSP (Malvern Instruments, Malvern, United Kingdom) with the 173° backscatter detector. Protein solutions were prepared in MES (final concentration of 25 μM), transferred to a DLS cuvette (precooled at 10°C), and quickly transferred into the Zetasizer. After incubation at 10°C, the first set of DLS measurements was conducted. Subsequent measurements were conducted at alternating temperatures (heating to 37°C and cooling to 10°C) for three cycles. The hydrodynamic diameter (Zavg) was calculated. Measurements were performed in triplicate.

In 20 mM MES buffer (pH 5.8), the FA leachate (Table 1) and 25 µM of RELP were mixed and kept at 4°C. The coacervation above the T _t was performed using a heat block at 37°C for 10 min, followed by centrifugation at 15,000 rpm at 37°C for 10 min, and the supernatant containing unbound metals was transferred to the new tube. Next, the RELP coacervates were resolubilized in phosphate–citrate buffer (4 mM Na₂HPO₄ and 100 mM citric acid, pH 2.2) at 4°C for 10 min for the desorption of bound REEs from RELPs. To separate the RELP and desorbed REEs, coacervation above T_t was triggered as described previously. The supernatant was recovered and used to analyze the recovered metals. The obtained RELP coacervates were resolubilized in fresh coal fly ash leachate for subsequent cycles (up to four cycles). The concentrations of the recovered REEs were determined using ICP-MS. The percentage of the metal recovered was calculated relative to the amount of metal added initially.

The RELP was incubated in four different solutions containing varying concentrations of REE (Tb³⁺) (1 µM to nM range) and equimolar concentrations of Mg and Zn (approximately 1 mM each). The REE was recovered as described above, except for the desorption of bound metal using citrate buffer (100 mM citric acid, pH 1.8). The concentrations of the recovered metal in the filtrate were analyzed using ICP-MS.

Metal ion purity, fold increase, and yield are defined as P u r i t y T b 3 + = C T b 3 + C T b 3 + + C m g 2 + + C z n 2 + , F o l d   i n c r e a s e   i n   p u r i t y T b 3 + = P u r i t y T b 3 + i n   r e c o v e r e d   s o l u t i o n P u r i t y T b 3 + i n   i n i t i a l   s o l u t i o n , Y i e l d = R e c o v e r e d   C T b 3 + I n i t i a l   C T b 3 +   .

The RELP used in this study is the LanM, derived from Methylorubrum extorquens AM1, fused to the C-terminus of the ELP. The generated de novo 3D structural model using AlphaFold2 (DeepMind) (Jumper et al., 2021) is shown in Figure 2A. As expected, the predicted structure showed an unstructured ELP region and the folded LanM protein region, suggesting that ELP fusion does not disrupt the folded structure of LanM. The RELP contains the ELP gene encoded with Val as the guest residue and comprises 150 repeats of the [GVGVP] amino acid motif (ELP[V150]). The RELPs purified through ITC have theoretical molecular weights of 75 kDa and were observed in the protein gel around the 75-kDa protein standard, with a purity above 95% (Figure 2B).

As discussed above, the phase transition of an ELP fusion protein is generally modulated by increasing the solution temperature above the inverse T_t, and by increasing the salt concentration, the transition is achieved even at temperatures much lower than T_t (Hassouneh et al., 2010). In the current study, we triggered the coacervation of the RELP by increasing the temperature of the purified protein solution above T_t, without adding additional salt. For this purpose, following purification, the T_t values of the RELP were measured by temperature-programmed turbidimetry. This technique monitors A_350nm of the RELP solution while the temperature increases. As T_t is concentration-dependent, T_t is characterized by a concentration series. The turbidity profile exhibits a sharp increase corresponding to the RELP phase transition from soluble to insoluble aggregates. The results showed that T_t of a RELP decreases as its concentration in solution increases (Figure 2C). The reversibility of the RELP phase transition is confirmed by measuring size changes as the ELP transitions from unimer to micron-scale aggregates below T_t and above T_t, respectively. The micrometer-scale coacervates of the RELP were observed at 37°C via DLS analysis. The DLS analysis also showed that the RELP maintains its reversible phase transition property during three cycles of temperature change. An insoluble coacervate is fully reversible upon cooling of the solution (when the solution temperature was decreased below T_t) and was completely resolubilized (Figure 2D). These results demonstrated that the LLPS of the RELP can be triggered by temperature only, and such a reversible transition of the RELP will improve the existing method for the repeated use of the RELP for selective recovery of REEs.

We investigated the REE recovery using the RELP from the coal fly ash leachate, an abundant and potentially valuable industrial REE feedstock. The acid leachate of FA (pH 4.5) in this study contained ≈152 mM of total metal ions (Table 1), including ≈5 µM REEs (0.003 mol% REEs, excluding monovalent ions) and other metals (4 mM Mg, 128 mM Na, and 19 mM Ca). Although the original coal fly ash leachate had Al, Si, and Fe, they were removed during pH adjustment. The RELP retained ∼95% of the initial REE binding capacity even after four cycles (Figure 3). The RELP selectively recovered the REEs for repeated cycles and showed minimal recovery for non-REEs, with no substantial decrease in its performance after four cycles of REE recovery from the FA leachate; the results were insignificant for each element between each cycle (Figure 3). The slight, apparent high recovery efficiency for light REEs (Nd and Sm) over heavy REEs (Figure 3) is likely due to the LanM, derived from M. extorquens AM1, favoring the larger and more abundant light REEs than heavy REEs (Gd–Yb) (Table 1), as previously reported (Deblonde et al., 2020; Dong et al., 2021).

In comparison, previous studies showed single-step selective recovery of REEs from complex industrial wastes (Hussain et al., 2022). In a single-step process, the RELP has shown high recovery efficiency for REEs (≈80%) from steel slag leachate containing 41 mM of metal ions (Hussain et al., 2022). Moreover, in another study, the incubation of LanM with leachates of lignite and electronic waste, followed by a single-size exclusion filtration step (using a centrifugal filter with a molecular weight cutoff), showed that all of the non-REE elements remained in the recovered filtrates, and REEs are selectively extracted by the protein fraction left in retentate (Deblonde et al., 2020). The retentates obtained from the filtration assay were digested and used for metal quantification by ICP-MS (Hemmann et al., 2023). Hence, regarding the reusability of biosorbents for the selective and repeated recovery of REEs from complex industrial wastes, this study outperformed comparable REE extraction processes using protein-based technologies (Deblonde et al., 2020; Hussain et al., 2022; Hemmann et al., 2023). Moreover, the method we used is entirely protein-based, and hence, it does not require any centrifugal filter to separate protein-bound REEs from unbound non-REEs or proteins from recovered REEs after the desorption step. Considering industrial applications, the simplistic non-chromatographic purification and high stability upon reuse of the RELP demonstrate the flexibility and potential for use as a low-cost recovery platform for REEs from complex waste sources.

REEs have been reported to exist at very low concentrations (picomolar scales) in various low-grade REE sources. We investigated the ability of the RELP to recover REEs from environmentally relevant concentrations (µM to nM range). The RELP was incubated in different solutions containing varying concentrations of REEs (Tb³⁺) (µM to nM range) and equimolar concentrations of Mg and Zn (∼1 mM each) at pH 5.8. Mg and Zn were selected as representative competing non-REE elements because REEs frequently coexist with these metals at high concentrations (high μM range) in ore and waste streams. The RELP selectively recovers the REE (Tb³⁺), which has a concentration as low as 1.8 nM (Table 2). Mattocks et al. constructed a LanM-based protein sensor. They confirmed its detection efficiency for REEs using fluorescence resonance energy transfer. The sensor detected REEs within a 10–50 µM range, with a 7-fold ratiometric response but only a weak response to divalent and trivalent metal ions at pH 7.0 (Mattocks et al., 2019). Moreover, in another study, Trp-substituted LanM directly quantified 3 ppb (18 nM) terbium in acid mine drainage using luminescence resonance energy transfer at pH 5.0 (Featherston et al., 2021). These studies anticipate that LanM can be used for detecting and quantifying REEs in extremely low concentrations in environmental and industrial samples. The results suggest the advantage of the RELP system for the selective recovery of high-purity REEs from extremely low-concentration metal solutions. We anticipate further applications of this system by re-engineering the RELP to selectively recover other metal ions (e.g., actinides) present at extremely low levels in seawater (Mattocks et al., 2022). Most previous biosorbent studies are more focused on environmental remediation than the extraction of REEs for industrial purposes, where the purity of the recovered metals is critical. The lanthanide-binding tag-engineered Escherichia coli cells showed REE purity of 0.11% in the extracted solution because of the adsorption of non-REEs to the cell surface, which eluted along with REEs (Brewer et al., 2019b). In terms of purity, LanM-immobilized magnetic nanoparticles showed 10.9 mol% REE purity in the recovered solution, which was 1,155-fold higher than in the FA leachate stock solution (0.009 mol%) (Ye et al., 2023a). The LanM-immobilized agarose microbead column showed 88.2 mol% REE purity, which was 2,040-fold higher than in the FA leachate stock solution (0.043 mol%) (Dong et al., 2021). The increase in REE purity by the RELP was 10–100 orders of magnitude higher than other benchmark materials, such as LBT-engineered cells and LanM-immobilized magnetic nanoparticles, and comparable to the LanM-based column. Notably, the fold increase in REE purity was also 10–100 order-of-magnitude higher than that of LanM-based biosorbents (Table 2). The results demonstrated that the RELP can be used to obtain high-purity REE mixture solutions by utilizing low-grade REE solutions. However, the RELP-based approach has limitations, such as the inability for intra-REE separation, as was demonstrated by the LanM-immobilized resin. We anticipate achieving even higher yields and product purity for other environmental and industrial waste leachates, which will be the subject of future studies.

It is noteworthy that the RELP platform can be leveraged to develop a more efficient biosorbent for recovering other critical elements and isotopes, provided it is re-engineered with other lanmodulin variants (Mattocks et al., 2022; Park et al., 2022).