HIGH EFFICIENCY, LOW ENVIRONMENTAL IMPACT BIO-PROCESS FOR PURIFICATION OF RARE EARTH ELEMENTS

CROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. provisional application no. 63/405,353, filed September 9, 2022, the entire disclosure of which is incorporated herein by reference.

RELATED INFORMATION

[0002] Rare Earth Elements (REE), typically referring to the lanthanides (lanthanum to lutetium) and sometimes scandium and yttrium, are essential ingredients for sustainable energy technologies including high strength lightweight magnets used in electric vehicles and wind turbines ^5,6 ; room temperature superconductors ⁷ ; lightweight high-strength alloys ^8,9 ; high-efficiency lighting ¹⁰ ; and battery anodes ¹¹ . All of these applications put an increasing demand on the global REE supply chain. As the world demand for sustainable energy grows ¹² , developing a sustainable supply chain for high-purity REE is critical ¹³ .

Many currently available methods for refining REE often involve harsh chemicals, high temperatures, high pressures, and generate a considerable amount of toxic waste ¹⁴ ' ¹⁶ . These processes give sustainable energy technologies reliant on REE a high environmental and carbon footprint.

[0003] The majority of REE chemical separations utilize commercially available organic solvents and extractants ¹⁷ . All lanthanides exist as trivalent cations and the ionic radius difference between the largest rare earth, La ³⁺ , and the smallest rare earth, Lu ³⁺ , is only 0.17 A ¹⁸ . This means that separations of adjacent or near-adjacent REE pose an enormous challenge for conventional chemical methods, requiring organic solvent extractions in extremely long mixer settler devices ¹⁹ . This results in large amounts of toxic waste being generated. As a consequence, due to its high environmental standards, the United States has no capacity to produce purified REE. Furthermore, only two REE purification plants exist outside of China ^15,16,20 .

[0004] Biomining is a promising alternative to conventional mining technologies, and already supplies 5% and 15% of the world’s gold and copper ²¹ . It is expected that a suitable REE-biomining system will operate in three steps: (1) bioleaching metals from an ore or end- of-life feedstock like a magnet; (2) separating the lanthanides from all other metals present in the leachate (e.g., uranium and thorium from an ore, or iron from a magnet); and (3) separating individual lanthanides. It is expected that the second and third steps will be carried out separately, removing the engineering challenge of simultaneously discriminating between lanthanide and non-lanthanide and individual lanthanides. Progress has been made in developing microorganisms for the first bioleaching step of the biomining process ^22,23 , and in the second total lanthanide separation step ²⁴ ' ²⁷ .

[0005] New biological and chemical methods ²⁸ ' ³¹ have recently been developed to address the challenges of total ²⁴ ' ^27,32 ' ³⁴ , light and heavy REE ³⁵ , and individual ^1,2,36 REE- separations. For example, lanmodulin, a REE-binding protein discovered in methylotrophic bacteria, is selective for REE in the presence of molar amounts of competing metal cations ^24,25 . Meanwhile, lanthanide binding tags (LBTs), attached either to the surface of Caulobacter crescentus ²⁶ or an engineered curb biofilm ²⁷ can selectively bind REE in the presence of competing metals. Finally, both Methylorubrum extorquens ²⁵ , and E. coli engineered with surface-displayed LBTs ³² are able to preferentially accumulate heavy lanthanides from a mixed solution of lanthanides. Despite numerous advances in total REE separation from competing metals, advancement of separation of individual REE from total REE remains a challenge. There is thus an ongoing and unmet need for improved approaches to REE separation from a variety of materials. The present disclosure is pertinent to this need.

BRIEF SUMMARY

[0006] Rare earth elements (REE) are essential ingredients of sustainable energy technologies, but separation of individual REE is one of the hardest problems in chemistry today ¹ . Biosorption, where molecules adsorb to the surface of biological materials, offers a sustainable alternative to environmentally harmful solvent extractions currently used for separation of rare earth elements (REE). The REE-biosorption capability of some microorganisms allows for REE separations that, under specialized conditions, are already competitive with solvent extractions ² , suggesting that genetic engineering could allow it to leapfrog existing technologies. In this disclosure, to identify targets for genomic improvement we screened 3,373 mutants from the whole genome knockout collection of the REE-biosorbing microorganism Shewanella oneidensis MR-1 ^3,4 . 130 genes were identified that increased biosorption of the middle REE europium, and 112 that reduced it. Biosorption changes from the screen for a mixed solution of three REE (La, Eu, Yb) were verified using Inductively Coupled Plasma Mass Spectrometry (ICP-MS) in solution conditions with a range of ionic strengths and REE concentrations. We identified 18 gene ontologies and 13 gene operons that make up systems that affect biosorption. The disclosure demonstrates that disruptions of a regulatory component of the arc system (hptA which regulates cellular response to anoxic environments and polysaccharide biosynthesis related genes wbpQ, wbnJ, SO 3183) consistently increase biosorption across all tested solution conditions. The largest total biosorption change comes from SO 4685, a capsular polysaccharide (CPS) synthesis gene, disruption of which results in an up to 79% increase in biosorption; and nusA, a transcriptional termination/anti-termination protein, disruption of which results in an up to 35% decrease in biosorption. Knockouts of glnA,pyrD, and SO 3183 produce small but significant increases (~ 1%) in relative biosorption affinity for ytterbium over lanthanum in multiple solution conditions tested, while many other genes we explored have more complex binding affinity changes. Modeling suggests that while these changes to lanthanide biosorption selectivity are small, they could already reduce the length of repeated enrichment process by up to 27%. The disclosure thus in part provides an analysis of how genetics affect REE-biosorption by S. oneidensis, and describe targets that can be manipulated to improve biosorption and separation of REE by genetic engineering. The disclosure includes modified microorganisms that are in the presence of one or more REEs. The modified organisms have a modification that affects expression of one or more described genes. In non-limiting examples the modification affects expression of one or a combination of SO 0456, wbpA, nusA, hptA, pyrD, SO 3099, etfB, wbpQ, SO 3183, wbpP, wzz, wzd, SO 3385, pyrC, arcA, mshQ, mshD, mshA, mshB, mshL, mshJ, pyre, SO 4685, and wbnJ. In non-limiting examples the modification comprises a disruption of SO 4685, a disruption of wbpA, a modification that results in increased expression of nusA, or any combination thereof. In embodiment, modification of any gene described herein can change the preference for a particular REE or type of REE exhibited by the modified bacteria.

BRIEF DESCRIPTION OF FIGURES

[0007] Arrows and lines are used in some figures to show to point to color features that are in not in black ink.

[0008] Figures 1A-1C. Screening the Shewanella oneidensis whole genome knockout collection finds 242 genes representing 18 gene ontologies that control Eu- biosorption. We used the Arsenazo III (As-III) competitive assay for europium- (Eu-) binding to screen 3,373 unique members of the S. oneidensis whole genome knockout collection to identify mutants with modified REE-biosorption capability. (A) Unbound As-III absorbance peaks at ~ 530 nm (resulting in a cyan color), while Eu-bound As-III (proposed structure) absorbance peaks at ~ 650 nm (purple). Right panel shows a computer-generated image of a sample assay plate derived from spectroscopic data. Higher biosorption by S. oneidensis results in a lower concentration of Eu- As-III and hence lower 650 nm absorption (the well will be more cyan-colored) while lower biosorption results in a higher concentration of Eu-As-III (the well will be more purple-colored). Additional information on the high- throughput screen is presented in Figure 5. (B) The As-III screen found 242 genes that control Eu-biosorption. The largest source of Eu-biosorption variability in the screen is due to bacterial density differences between mutants. For most mutants, the optical density of the culture at the start of the biosorption screen will map onto As-III absorption at 650 nm by a linear piecewise function (shown as a blue solid line). Mutants shown as red diagonal crosses had significantly less biosorption than the plate average. Mutants shown as green horizontal crosses had significantly higher biosorption than the plate average (mutants shown as blue dots are not significantly different from the average). (C) Gene ontology enrichment analysis found that 18 ontologies were enriched with mutants discovered by the As-III screen. The dotted line indicates a - value of 0.05. We only show results with /?-values below 0.05 and gene ontologies with > 1 representative mutant. Numbers above each bar indicate the number of significant biosorption genes within each ontology in the screen results relative to the number in the S. oneidensis genome. IMP: inosine 5 '-monophosphate; UDP-GlcNAc 4- epimerase: UDP-N-acetylglucosamine 4-epimerase; Ubi-cyt-c reductase: ubiquinol- cytochrome-c reductase.

[0009] Figures 2A-2F. Operon enrichment, ontology enrichment, and regulatory analyses pinpoint 6 groups of genes that influence multiple mechanisms behind Eu- biosorption by S. oneidensis. (A to C) The results of the high-throughput Eu-biosorption screen of the S. oneidensis knockout collection were analyzed to find operons with statistically-significant enrichments of hits. The location of the transposon disruption in each gene is marked as a black line. Here we show three operons that are the most statistically- significant results of this analysis. (D) The pyrimidine synthesis pathway was selected by ontology enrichment analysis. (E) One gene involved in the Anaerobic Redox Control (Arc) regulatory system (hptA), as well as two genes regulated by Arc whose knockouts produced differential biosorption were also selected for further analysis. (F) Finally, five genes whose knockouts produced some of the largest changes to Eu-biosorption were also selected for further analysis.

[0010] Figures 3A-3H. ICP-MS measurements validate the results of high- throughput Eu-biosorption screening in up to 79% of cases. Bar plots show levels of lanthanum (blue), europium (yellow), ytterbium (red), and total REE (grey) biosorption for each strain. The error bar indicates the standard deviation on the total biosorption of three biological replicates. The number of stars above each bar indicates the statistical significance of the measurement difference from quasi-wild-type (A to D) and wild-type (E to H): *: p- value < 0.05; **: /?-value < 0.01; ***: /?-value < 0.001. 5 indicates a transposon insertion mutant (in panels A to D), while A indicates a clean deletion mutant (in panels E to H). Note the nSO 2183 mutant which indicates that the transposon is near to, but not within SO 2183. Cross-checks of As-III Eu-biosorption assay and ICP-MS measurements with transposon mutants are shown in Table S2. (A) The low ionic strength, low initial REE concentration environment matches 53% of the As-III screen (Table S2). (B) The low ionic strength, high initial REE concentration environment (LH) recapitulates the highest percentage (63%) of results of the As-III screen. (C) The high ionic strength, low REE environment (HL) reproduces 63% of significant changes to biosorption. (D) The high ionic strength, high initial REE environment reproduces the smallest number (42%) of results from the As-III screen. (E to H). Clean deletion mutants replicated at least some of the results of transposon mutant measurements in three of four cases.

[0011] Figures 4A-4E. Nine gene disruption mutants make notable changes to REE-biosorption selectivity. (A to C) For most transposon insertion mutants (including those with modified total REE biosorption), under most of the four solution conditions tested, individual REE biosorption is linearly related to total biosorption or individual biosorption of either of the other 2 REE tested (grey circles, and the black dashed fit lines in panels A to C) over a finite range of REE biosorption (note the finite extend of dashed black lines in panels A to C). (A to D) Individual points indicate the mean values of the mutants and the error bars show the standard deviation along the axis of maximal variation. (A to C) We highlight changes in La and Yb affinity for 6pyrD in the LH environment as well as two mutants (brnisA and dSO 4685) who’s total biosorption was too small (OnisA) or large (dSO 4685) to compare to our finite line of best fit, yet nonetheless clearly had different La and Yb affinity relative to the other mutants. In particular, note how both brmsA and dSO 4685 have similar La biosorption to other mutants in (C) yet had very different Yb biosorption. (D) We had insufficient data to perform a line of best fit on our clean gene deletion data, yet it is clear from this plot that the glnA deletion has an increase in relative Yb/La affinity. (E) Here, we display all our mutants with robust biosorption changes (mutants that enhanced or decreased relative biosorption in multiple environments, or when a particular REE had enhanced or lowered relative biosorption in more than half the genes in a particular group of interest). The number of stars above or below each bar indicates the statistical significance of the measurement difference from quasi-wild-type: *: -value < 0.05; **: /?-value < 0.01; ***: p- value < 0.001. 5 indicates a transposon insertion mutant, A indicates a clean deletion mutant. Error bars indicate standard deviation of three biological replicates. PSI and 2: Polysaccharide Synthesis 1 and 2; MSHA: MSHA Pilus Assembly; Pyr: Pyrimidine Synthesis.

[0012] Figure 5. Absorbance measurements were used to perform quality-control tests on the Arsenazo III screen for differential Eu-biosorption.

[0013] Figures 6A-6M. Thirteen operons are significantly enriched in genes influencing biosorption.

[0014] Figure 7. REE-biosorption and separation factor appear to equilibrate during the incubation time used in measurements in this study.

[0015] Figure 8. REE-biosorption by quasi-wild-type strains of S. oneidensis is lower than the true wild-type.

[0016] Figure 9. Recombined wild-type S. oneidensis strains do not have significantly different biosorption compared to the original wild-type.

[0017] Figures 10A-10D. ICP-MS measurements find 23 transposon insertion mutants of S. oneidensis with statistically significant changes in relative REE-biosorption in at least one solution environment, although few of these changes are robust.

[0018] Figure 11. Proposed schemes for REE separation by biosorption and desorption.

[0019] Figure 12. Effect of changes to REE-biosorption selectivity on REE- separation with a single-site-type binding model.

DETAILED DESCRIPTION

[0020] Unless defined otherwise herein, all technical and scientific terms used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0021] Every numerical range given throughout this specification includes its upper and lower values, as well as every narrower numerical range that falls within it, as if such narrower numerical ranges were all expressly written herein.

[0022] The disclosure includes all polynucleotide and amino acid sequences described herein. Each RNA sequence includes its DNA equivalent, and each DNA sequence includes its RNA equivalent. Complementary and anti-parallel polynucleotide sequences are included. Every DNA and RNA sequence encoding polypeptides disclosed herein is encompassed by this disclosure. Amino acids of all protein sequences and all polynucleotide sequences encoding them are also included, including but not limited to sequences included by way of sequence alignments. Sequences of from 80.00%-99.99% identical to any sequence (amino acids and nucleotide sequences) of this disclosure are included.

[0023] The disclosure includes all polynucleotide and all amino acid sequences that are identified herein by way of a database entry. Such sequences are incorporated herein as they exist in the database on the effective filing date of this application or patent.

[0024] The disclosure includes modified microorganisms having any modified single gene, and modifications of all combinations of genes described herein in the text, figures, figure legends, and tables of this disclosure.

[0025] Any gene and any combination of the genes that are described herein may be excluded from the claims of this disclosure. In embodiments, a modified microorganism of the disclosure may comprise or consist of only one modification of a single gene. In embodiments, a modified microorganism of the disclosure may comprise or consist of any combination of gene modifications described herein. In embodiments, only one or only a combination of genes that influence biosorption are modified.

[0026] The disclosure includes modifications that disrupt one or a combination of genes, modifications that increase expression of one or a combination of genes, or a combination of modifications that decrease expression of one or more genes and modifications that increase expression of one or more genes. Thus, the modifications involve altering the expression of one or more genes. Increasing, e.g., overexpressing a gene, can be achieved using various techniques that will be apparent to those skilled in the art when given the benefit of the present disclosure. In embodiments, increasing expression of a gene is achieved by substituting an endogenous promoter with a promoter that increases expression of the gene, relative to expression of the gene that is produced by the endogenous promoter. By “substituting” a promoter it is meant that the endogenous promoter (e.g., the promoter that is ordinarily operatively linked to the gene of interest without genetic engineering) has been changed so that is does not drive expression of the gene in the modified bacteria, and therefore the substituted promoter drives gene expression. By making this change, more mRNA is transcribed, thus facilitating production of more protein encoded by the pertinent gene that is operatively linked to the promoter. Substituting a promoter can include inserting a new promoter, while leaving the endogenous promoter in place, or inserting the new promoter in place of the endogenous promoter. The promoter that is inserted so that it is operably linked to and therefore drives expression of the described gene(s) can be heterologous to the bacteria, meaning it is taken or derived from a different organism, or it may be endogenous to the organism but has been introduced into a new location such that it can drive expression of the described gene(s). Various prokaryotic promoters that are suitable for this purpose are known in the art and include, for example, tufa and tufB. The substituted promoter (e.g., the promoter that is introduced into the bacteria) may be a constitutive or inducible promoter. The substituted promoter may be a core promoter, a proximal promoter, or a distal promoter. As an alternative or in addition to promoter modification, the disclosure includes addition of and/or repositioning of enhancer elements to increase expression of the described gene(s).

[0027] As an alternative or in addition to changing promoters, the disclosure includes increasing copy number of the gene that is to be overexpressed. In embodiments, one or more copies of the gene can be inserted into a bacterial chromosome, or can be introduced into bacteria using a plasmid. The additional copies of the gene may be in tandem, such as in a polycistronic configuration, or may be separated by segments of the bacterial chromosome or plasmid. In embodiments, a composition comprising the described bacteria are modified by transformation using one or more plasmids, which may be configured to be replicated and transferred to other bacteria in a bacterial population, such as by horizontal transfer.

[0028] In another embodiment, the disclosure comprises decreasing expression of genes. Decreasing expression can be achieved using any suitable approach. In embodiments, decreasing expression comprises disrupting the gene such that the protein encoded by the gene is not produced, or a protein produced by the gene does not function in the same way as if it had not been modified. In embodiments, a protein that is encoded by a modified gene of this disclosure is produced but does not function to impede bioleaching of REEs from a composition comprising them. In embodiments, a modification of a gene comprises a knockout of some or all of the gene. Modifications of the genes can be achieved using any suitable genetic engineering techniques. In non-limiting embodiments, the modification comprises an insertion, a deletion, or a combination thereof. In an embodiment, the modification comprises a complete knockout of the gene. An insertion, a deletion, and a complete knockout of the gene each comprise a disruption of the gene. The disclosure includes insertion within, or a deletion of any segment of a gene, including but not limited to a insertion or deletion of a single nucleotide, such that the encoded protein is not produced or its function is eliminated or reduced. In embodiments, an insertion replaces some or all of the described gene(s). In a non-limiting embodiment, the described gene(s) is modified by insertion of a transposable element. In non-limiting embodiments, the genes are modified using compositions and methods described in U.S. patent 11,053,493, from which the entire description is incorporated herein by reference. In embodiments, a modification of a gene comprises an insertion as described in Anzai, Isao A., et al. “Rapid curation of gene disruption collections using Knockout Sudoku." Nature Protocols 12.10: 2110-2137 (2017), from which the entire disclosure is incorporated herein by reference. In alternative approaches, site specific nuclease, such as Cas nucleases, can be used to modify any of the described genes. In embodiments, a type I, type II or type III CRISPR system can be used. Thus, in embodiments, a guide-RNA directed nuclease can make any of the described modifications. In embodiments, recombination of a chromosome or plasmid can be used, such as by introducing a recombination template comprising additional copies of a gene, and/or a promoter, to facilitate recombination of the recombination template into a desired location. In an embodiment, homologous recombination is used, and as such, the recombination template includes left and right homology arms to specify the location of recombination. In embodiments, a transposon system can be used to interrupt a gene sequence, such as the Sleeping Beauty transposon system.

[0029] In embodiments of this disclosure, one or a combination of genetic modifications within a microorganism are made to provide a modified microorganism. In embodiments, at least one engineered genetic change affects expression of one or a combination of SO 0456, wbpA, nusA, hptA, pyrD, SO 3099, etfB, wbpQ, SO 3183, wbpP, wzz, wzd, SO 3385, pyrC, arcA, mshQ, mshD, mshA, mshB, mshL, mshJ, pyre, SO 4685, and wbnJ. In one embodiment the at least one engineered genetic change comprises a disruption of SO 4685, a disruption of wbpA, or a combination thereof. In an embodiment, the at least one genetic change consists of a disruption of SO 4685. In an embodiment, the at least one genetic change consists of a disruption of wbpA.

[0030] In an embodiment, the at least one genetic change consists of increased expression of nusA.

[0031] Any one or a combination of these modifications may result in an increase in biosorption of one or a combination of Rare Earth Elements relative to biosorption exhibited by unmodified cells, or relative to cells that have different modifications.

[0032] The disclosure includes compositions comprising one or more REEs and modified bacteria of the disclosure. As such, the modified bacteria may be provided in combination with only one, or more than one REE, representative examples of which include but are not limited to light, middle and heavy REEs. In embodiments these REEs are La, representing light REE; Eu, as an example of a middle REE, and Yb, representing a heavy REE. [0033] The disclosure includes a biolixiviant produced by the modified bacteria and one or more REEs. In embodiments, the disclosure relates to separating combinations of REEs. The composition comprising the REEs may be any composition of matter, including but not limited to solids, semi-solids, and liquids. In embodiments, the REEs are present in a feedstock. In non-limiting embodiments, the REEs are present in coal fly ash, virgin ore, electronic waste, fluid cracking catalysts, and the like.

[0034] The disclosure includes a method comprising contacting a composition comprising one or more types of REEs with modified bacteria and/or a biolixiviant produced by modified bacteria of this disclosure. In an embodiment, the method further comprising separating and optionally purifying one or more types of REEs from the composition comprising the REEs and the biolixiviant.

[0035] In embodiments, a method of the disclosure comprises exposing a material and/or a fluid that comprises REEs to described modified bacteria to provide a mixture comprising the modified bacteria and the REEs, separating one or more REEs from the mixture to obtain a first composition comprising separated REEs, and repeating the process using the first composition that comprises separated REEs to increase separation and/or purity of separated REEs. Thus, the disclosure provides for iterative processing of compositions that comprise REEs to increase the amount and/or purity of separated REEs, such as by sequential biosorption and desorption from the surfaces of described modified bacteria.

[0036] A representative list of genes for which expression can be modulated is provided in Table A. In Table A, in the column Higher or Lower Biosorption, “Higher” means that separation of REEs is improved when the gene is disrupted. “Lower” means that increasing expression of the gene should result in improved REE separation.

Table A.

[0037] The disclosure comprises isolated modified bacteria, cell cultures comprising the modified bacteria, and kits comprising the modified bacteria. In an embodiment, a kit comprises one or more sealed containers comprising the modified bacteria, which can be used in REE bioleaching approaches.

[0038] The disclosure includes media in which the bacteria are cultured, and bacterial secretions. In an embodiment, the disclosure provides a biolixiviant produced by the described bacteria. In embodiments, the kit contains a sealable or sealed container that contains a biolixiviant produced by the described bacteria. The disclosure also includes modifying bacteria so that they comprise at least one of the described gene modifications. [0039] The disclosure includes all modified microorganisms described herein. The described approaches may be used to engineer any type of bacteria. In embodiments, the bacteria are Gram-negative bacteria. In embodiments, the bacteria are obligate aerobes. In an embodiment, the modified microorganisms are at type of Shewanella. In embodiments, the Shewanella is S. oneidensis. In embodiments, modified S. oneidensis preferentially adsorb (and possibly desorb as indicated by the disclosure) certain REEs over non REEs. In embodiments, the modified Shewanella preferentially adsorb (or desorb in certain conditions as indicated by the disclosure) particular Rare Earth Elements over other Rare Earth Elements. The disclosure includes making and using multiple strains of modified Shewanella for separating out different Rare Earth Elements. In embodiments, the disclosure provides combinations of modified and unmodified bacteria for use in the described methods. The disclosure includes using the modified bacteria in high ionic strength and low ionic strength environments. [0040] With respect to the present disclosure, biosorption and desorption from the surface of a microbial cell offers an environmentally-friendly route for individual REE- separation. Biosorption can provide high metal binding capacity a low cost ^37,38 . The cell surface, containing proteins ³⁹ , lipids ⁴⁰ , and polysaccharides ⁴¹ , offers a rich chemical environment for selectively binding and releasing REE. The membranes of both gramnegative and gram-positive bacteria contain sites that bind REE ⁴² .

[0041] The present disclosure demonstrates that by mutagenizing genomic loci (particularly regulatory regions, but also small sections of protein coding sequences) involved in REE biosorption, REE-biosorption can be engineered selectivity. The disclosure provides comprehensive profiling of the genetics of REE biosorption in S. oneidensis. Modulation of gene expression and related effects are demonstrated in this are demonstrated using S. oneidensis.

[0042] This disclosure includes the results of a high-throughput genetic screen to identify genes involved in biosorption of Eu. As this screen contains numerous possible sources of systematic (variations in growth phase, and bacterial density across each 96-well plate) and random noise (variations in reagent concentration, inoculum quantity, and crosscontamination due to splatter) we conducted a global analysis of the genetic screen results to identify which gene categories were statistically enriched with hits from the described screen to identify which biological mechanisms truly contributed to biosorption. The disclosure describes genes of interest and their validation, and includes description of how total biosorption and REE specificity changes across several different solution conditions, as demonstrated by the following Examples which are not meant to limit the disclosure.

[0043] Genetic Screen Finds 242 Genes that Influence Europium Biosorption [0044] We screened 3,373 unique members of the S. oneidensis whole genome knockout collection ^3,4 for genes that control biosorption of the middle REE europium (Eu) using an Arsenazo III (As-III) colorimetric screen ⁵⁴ (Figures 1A and 5, Materials and Methods). We hypothesized that given (1) the tendency of the affinity of many bacterial REE binding sites to have a preference for middle REE or (2) monotonically change with atomic number ²⁴ (e.g., the site will bind heavy REE more strongly than middle REE, and middle more strongly than light REE), and (3) that no lanthanide-binding site is likely to be totally selective for an individual lanthanide, that the middle-REE Eu is most likely to bind to most sites. As a result, screening for mutants with altered Eu-biosorption would reveal the largest number of mutants with both altered total biosorption and altered selective biosorption. For example, knocking out a site with no selective preference for lanthanides will reduce Eu-, La- , and Yb-binding equally. At the same time, because it is unlikely that any site is perfectly selective, the position of Eu in the lanthanide series means that an assay using it will detect more changes due to disruption of selective sites. For example, knocking out a site with high La-affinity will still reduce Eu-binding somewhat, but might not impact Yb-binding as much. Likewise, knocking out a site with high Yb-affinity will also reduce Eu-binding, and is likely to reduce La-binding even less.

[0045] In total, we found 130 gene disruption mutants that have significantly higher Eu-biosorption, and 112 that have significantly lower Eu-biosorption (Figure IB). We note that there was a slight (2.9%), but statistically significant (p < .0001), decrease in the average growth OD of our biosorption outliers compared to the non-outliers. This suggests that mutations that induce biosorption changes are more likely to have a small growth defect compared with mutations that do not.

[0046] Confirming Replicability of Genetic Screen Results

[0047] To test the efficacy of the described screening procedure, we randomly chose 64 mutants from 7 plates (18 high biosorption outliers, 14 low biosorption outliers, and 32 non-outliers), colony purified them, and conducted 3 biological replicates following similar procedures to our original screen and having each replicate in a new plate position. 62 of our mutants of interest grew enough to be able to analyze the data. We found that 53% (9/17) high biosorption outliers recapitulated the original screen and 31% (4/13) of low biosorption outliers. In total, 43% of the outliers recapitulated (13/30). 97% (31/32) of the non-outliers from the original screen remained non-outliers.

[0048] 18 Gene Ontologies are Significantly Enriched Among Genes Influencing

REE Biosorption

[0049] Using Fisher’s exact test, we analyzed each set of gene disruption mutants — those with higher or lower Eu-biosorption — for enrichment of gene ontologies, as defined by the Gene Ontology Consortium ^55,56 to identify trends in the overall set of genes contributing to biosorption. We identified 18 gene ontologies that were significantly enriched (p < .05) and had more than one representative gene within our genetic screen results (Figure 1C, Materials and Methods). Ten were enriched among genes whose disruption increases biosorption while eight were enriched among genes whose disruption decreases biosorption. Ontologies discussed in this disclosure whose gene disruptions increase Eu-biosorption include pyrimidine synthesis, polysaccharide synthesis, and histidine kinase activity ontologies, while one ontology discussed in this disclosure whose gene disruptions decrease Eu-biosorption is pilus assembly. [0050] 13 Operons are Significantly Enriched in Genes Influencing REE

Biosorption

[0051] Using computationally produced predictions for operons in S. oneidensis ^{51 58} and Fisher’s exact test, we identified thirteen operons with statistically significant (p < 0.05) enrichments of genes whose disruptions produced differential biosorption in the Arsenazo-III screen (Figures 6A-6M, Materials and Methods). Out of the thirteen, three operons had highly significant enrichment of hits (p < .001). Identifying enriched operons allowed us to further refine analysis of which functions contribute to biosorption. For example, while ‘polysaccharide synthesis’ is highlighted as relevant from the ontology analysis, the operon analysis points to two specific polysaccharide and O-antigen synthesis operons (‘PSI’ and ‘PS2’) that impact biosorption (Figures 2A and B). The operon analysis also specifies the MSHA pilus — previously discussed herein as more generally in the ‘pilus assembly’ ontology in the ontology enrichment analysis — as important to REE biosorption as 12 out of the 15 gene disruptions in this operon produced lower Eu-biosorption in the screen.

[0052] Regulatory Analysis of the Arc System Highlights Several Other Genes Important to Biosorption

[0053] A disruption in the histidine kinase hptA was found to substantially increase biosorption in the genetic screen and hptA contributes to the enrichment of the ‘histidine kinase’ ontology group among hits. Since HptA is a regulatory protein, we analyzed its impact on biosorption to determine if it is caused by indirect activation or repression of downstream genes and thus sought to identify the genes responsible.

[0054] HptA is part of the two-component Anoxic Redox Control or Aerobic Respiration Control (Arc) system in S. oneidensis. ArcS phosphorylates HptA in the absence of O2, and HptA in turn phosphorylates the response regulator ArcA ^{59 60} . While we did not have gene disruptions we could screen for arcS or arcA we analyzed whether disruption of hptA produces a similar effect as disruption of arcS. We found that individual disruption of 29 of the 604 genes whose activity is affected by an arcS deletion (and thus likely a hptA deletion as well) ⁵⁹ produced significant changes in Eu-biosorption (Table SI).

[0055] Of the ArcS -regulated genes that may contribute to biosorption, several disruptions decreased biosorption (such as betfB, a disruption of subunit B of the e' transfer flavoprotein ⁶¹ ); several increased biosorption (such as 5 SO 3099, a disruption of an outer membrane long-chain fatty acid receptor ⁶¹ ); and several were also found to contribute to enriched ontologies (such as dpyrE which is involved in pyrimidine biosynthesis ⁶¹ ). (‘5’ indicates a gene disruption mutant and ‘A’ indicates a gene deletion mutant. [0056] Six Groups of Genes that Influence Multiple Mechanisms of REE Biosorption were Chosen for Detailed Analysis

[0057] Our As-III biosorption screen could analyze only a single REE. We thus sought to expand our analysis by looking at how our gene disruptions impacted biosorption of multiple REE. We selected six groups of genes representing a wide range of cellular functions for detailed analysis with the Inductively Coupled Plasma Mass Spectrometry (ICP- MS) — an instrument for making robust measurements of concentrations of multiple elements — in order to validate the results we found with the As-III assay. We focused on groups of genes rather than individual outliers to improve on an n = 1 screen. Finding multiple genes with similar functions or with similar locations in the genome that impacted biosorption supported the interpretation that these genes were truly significant. We selected these groups based on gene ontology enrichment, operon enrichment, and regulatory analyses from our Eu-screen biosorption data.

[0058] For three of these groups, we selected operons of interest: polysaccharide and O-antigen synthesis operons 1 and 2 (PSI and PS2; Figures 2A and B) and the MSHA pilus operon (MSHA; Figure 2C). Within the MSHA operon, we chose to look at <3mshQ, <3mshl), bmshC. bmshA. bmsh . bmshl.. bmsh.J because these are all predicted to be either outer membrane proteins or found on the pilus appendages ⁶³ . All of these, except for 6mshC, had significantly lower biosorption in the As-III Eu-biosorption screen.

[0059] An additional group of disruptions in non-contiguous genes (dpyrC, dpyrD, and 6pyrE) was selected for detailed study based on their contribution to the enrichment of the pyrimidine biosynthesis gene ontology (Pyr; Figure 2D).

[0060] In addition to genes contributing to ontology and operon enrichment, we chose to analyze disruptions in genes from the Arc system, including a disruption of the Arc system histidine kinase dhplA as well as disruptions of two genes regulated by the Arc system: detfB and 3SO 3099 (Arc; Figure 2E).

[0061] As a miscellaneous group, we selected five gene disruption mutants that were independent of any identified grouping but produced strong changes in biosorption (Figure 2F). A mutant in the gene coding for the transcriptional termination/anti-termination protein NusA (8/w.sd) ⁶¹ (also see Figure IB, left panel) had lower biosorption. Meanwhile disruption of SO 4685, which codes for a protein involved in extracellular capsular polysaccharide synthesis (CPS) ⁶¹ produced higher biosorption (also see Figure IB, right panel). Likewise, disruption of SO 3385 which codes for a transcriptional activator of singlet oxygen protection ⁶¹ produced higher biosorption. Finally, an insertion 150 bp upstream of SO 2183, which codes for a protein involved in biosynthesis of the cell wall component peptidoglycan ⁶¹ , had higher biosorption.

[0062] ICP-MS Validates Differential Biosorption of Genes from Selected Gene Groups

[0063] ICP-MS measurements of mixed REE-biosorption largely validated the results of the high-throughput biosorption assay (Figures 3A to D, Table S2). We measured biosorption of lanthanum (La; representing light REE), Eu (for middle REE), and ytterbium (Yb; for heavy REE) under four solution conditions (Materials and Methods) by 25 gene disruption mutants representing the six groups of genes chosen for detailed analysis (Figures 3A to D) and six clean deletion mutants (Figures 3E to H).

[0064] In industrial settings, REE are processed in a wide array of combinations and concentrations, and with a wide range of competing metal concentrations. We chose to explore two axes of interest to characterize our selected insertion mutants: low and high ionic strength (provided by sodium chloride); and (2) low and high total initial REE concentration. We tested multiple ionic strength concentrations to investigate the possibility of cation competition for REE binding sites and the ability of ionic strength to change protein configurations among other potential impacts. The disclosure thus includes use of four solution conditions: low ionic, low initial REE concentration (LL); low ionic, high REE (LH); high ionic, low REE (HL); and high ionic, high REE (HH). Biosorption solution conditions used are detailed in Table 1.

[0065] As a benchmark for comparison, we selected 4 quasi-wild-type (qWT) transposon insertion mutants from the S. oneidensis whole genome knockout collection that had the insertion in a location unlikely to impact biosorption (Materials and Methods). We found that the genuine wild-type S. oneidensis showed at least 13% higher biosorption than the average qWT in every solution condition (Figure 8), indicating that the presence of a transposon insertion alone may affect biosorption. As we were interested in how knocking out the gene qualitatively affected biosorption rather than the absolute change from the wildtype, we thus chose to compare certain transposon mutants with the average of our qWT mutants rather than the true wild-type.

[0066] The multiple-REE biosorption assay recapitulated significant increases or decreases in biosorption from the As-III screen for 54% (13/24) of disruption mutants tested for the LL environment (Figure 3A), 62% (15/24) for the LH (Figure 3B) and HL (Figure 3C) environments, and 42% (10/24) for the HH (Figure 3D) environment. 79% (19/24 genes) of our insertions are validated in at least one environment (Figures 3A-H and Table S2). Recapitulation of most of our results in at least some conditions supports validation of the initial screen. Furthermore, the percentage of insertions chosen by our operon and ontology enrichment method that were validated in at least one environment is much higher than the percentage of randomly chosen insertions that replicate (79% vs. 43%). This suggests that the described enrichment analysis approach does improve the chances of finding insertions that produce robust changes to biosorption.

[0067] Disruptions to 9 genes produced higher biosorption in all environments: disruption of the uncharacterized protein Wzd (bwzd) (Polysaccharide Synthesis Operon 1), disruption mutants for all 3 of the chosen Polysaccharide Synthesis Operon 1 genes, disruptions to all 3 genes related to the Arc system (including 6etfB, which had lower biosorption in the genetic screen), and an insertion upstream of the LD-transpeptidase encoding gene SO 2183 and disruption to the CPS-synthesis gene SO 4685 from the Miscellaneous group. In fact, dSO 4685 produced our largest observed increases in total biosorption ranging between 31% (in LL) and 79% (in HH) higher than the qWT.

[0068] The insertions for SO 0456 and SO 3385 (both in the Miscellaneous group) had significantly higher biosorption than the qWT in every solution condition except for HH. [0069] Only one gene disruption produced consistently lower total REE-biosorption. Specifically, disruption of the last 10% of the coding region for the transcriptional termination/anti-termination protein NusA (6////.sd; the Miscellaneous group), produced our largest observed reductions in total biosorption ranging between 11% (in LL) and 35% (in HH) lower than the qWT. Thus, the disclosure includes upregulating this gene.

[0070] Five gene disruption mutants showed a notable discrepancy in total biosorption between low ionic strength (Figures 3A, B, E, and F) and high ionic strength (Figures 3C, D, G, and H) environments. For example, the MSHA genes showed the greatest environment dependency in their biosorption results. Additionally, 6mshJ had significantly lower biosorption in the high ionic strength cases (HL and HH; Figures 3C and D), but no significant change in the low ionic strength cases. Meanwhile, 6mshA and 6mshB (MSHA Pilus had lower biosorption capabilities in the low ionic strength environments (LL and LH; Figures 3A and B), and either no significant change or a biosorption increase in the high ionic strength environments.

[0071] All three pyrimidine biosynthesis gene insertions had significantly higher biosorption (between 7 and 11% higher than the qWT) in the HL environment but registered no significant difference in any other environment (Figure 3D). [0072] Clean Gene Deletion Mutants Largely Verify Biosorption Results for Gene Disruption Mutants

[0073] We created clean deletion mutants for four genes whose disruption conferred standout biosorption changes: ^msh. J. ShptA, ASO 3385, S.SO 4685. Three of four of these clean deletion mutants at least partially reproduced the results of the corresponding disruption mutants. While insertion mutants are effective at knocking out gene function, they do not always successfully mimic a true single gene knockout. While the S. oneidensis knockout collection was designed to mitigate polar effects ^53,64 ’ ⁶⁵ , it is possible that the early parts of the knockout gene can still create a partially functional product ⁶⁶ , which could account for the discrepancies between deletion and disruption strains. We also created two additional mutants for genes identified by the Arsenazo III genome-wide screen : 8.80 0625 (a knockout for periplasmic cytochrome c oxidase regulatory protein) and tSglnA (for glutamine synthetase).

[0074] Clean deletion of the pilus biogenesis gene mshJ (Smsh.J) had significantly lower biosorption (11 to 17%) in three of the four environments tested (all except HL where the biosorption level was lower, just not statistically significantly). The main difference between the insertion and the clean deletion mutant is that 8msh.J produces lower biosorption in the low ionic strength environments, while 6mshJ has no significant change compared to the qWT in those environments.

[0075] Clean deletion of the transcriptional regulator gene hptA (8hptA ) produced large increases in biosorption (16 to 39%) in high ionic strength environments (HL and HH; Figures 3G and 3H) just like the transposon mutant (6hptA) (Figures 3C and 3D). However, unlike the transposon mutant, ShptA did not produce significantly different biosorption from the wild-type in low ionic strength environments (LL and LH; Figures 3E and F).

[0076] Complete knockout of the capsular EPS biosynthesis gene SO 4685

(8SO 4685) showed a significant increase in biosorption (15 to 40%) in every environment just like the insertion mutant (Figures 3A-H).

[0077] Clean deletion mutants for SO 0625 and glnA produced significantly lower biosorption than the wild-type in the low ionic strength cases, and non-statistically significantly lower biosorption in the high ionic strength cases.

[0078] Nine Insertion Mutants Have Notable Modification of Individual Lanthanide Binding Preference

[0079] We next examined if, in addition to changes to total REE biosorption, our mutants produced changes to the relative biosorption affinity for particular REE over others. Out of the 25 insertion mutants we conducted follow up ICP-MS analysis on, nine of our mutants appeared to produce robust changes to biosorption preferences for particular REE. [0080] Our qWT already has a marked preference for heavier REE. This preference for heavier REE increases with ionic strength. From an initially equimolar mixed REE solution of La, Eu, and Yb, the qWT-biosorbed fraction contains between ~ 19% and ~ 28% La; ~ 37% and 40% Eu; and ~ 35% and 44% Yb (Figure 7).

[0081] For most transposon insertion mutants, under most of the four solution conditions tested, individual REE biosorption is linearly related to total biosorption (Figure 4, Materials and Methods) over a finite range (note the finite extent of dashed black lines in Figures 4A to B).

[0082] However, every transposon insertion mutant that we tested had at least one REE biosorption result in at least one environment that deviated from the linear individual-to- total relationships established for most mutants under most conditions (Figures 4A to 4C, and 8). To improve statistical confidence we narrowed our criteria to select for ‘robust’ results by focusing on only those mutants with enhanced (or decreased) relative biosorption of a particular REE in the same direction in multiple environments or when a particular REE had enhanced or lowered relative biosorption (again, in the same direction) in more than half the genes in a particular group of interest.

[0083] Seven mutants produced robust results according to our criteria. These highlighted mutants are summarized in Figure 4E.

[0084] Disruption of genes in the Polysaccharide Synthesis 1 operon tend to increase Yb-binding and decrease La-binding for high ionic strength environments (Figure 4E, Figures 10A-D). For example, disruption of SO 3183 increases relative Yb-binding under HH by 2.2% and reduces relative La-binding by 4.1%.

[0085] Disruption of two of the four genes in the Polysaccharide Synthesis 2 operon (wzz and wbpA) produces a significant increase in Eu and a significant decrease in Yb biosorption in the LH condition (Figures 10A-D). The disruption of wbpA has a significant relative increase in binding of Eu under the LL condition as well.

Disruption of mshB ( rnshB) produces significant reductions of in La-binding in low ionic strength conditions, likely at the expense of Eu-binding (although these changes are not significant). [0086] Disruption of pyrD produces significant increases in Yb-binding coupled to reductions in La- and Eu-binding in both LL and LH. Under LH, the increase in Yb-binding of 2.9% is one of the largest significant increases in relative REE binding.

[0087] While disruption of pyrC under HH produces the largest significant change, increasing La-binding by 5.8%, it did not meet our robustness metric.

[0088] Two insertion mutants from the Miscellaneous group (bnusA and dSO 4685) have total biosorption levels that are so different from the rest of our insertion mutants that we did not include them in our formal analysis of relative REE changes (they were outside of the finite linear fit region in Figure 4A-C). However, in the case of the LH environment, we still found a clear way of illuminating relative REE affinity changes (Figure 4C). bnusA produces very similar La-binding to other transposon insertion mutants that were in-range for our analysis. At the same time, it had a much lower relative level of Yb biosorption. This made it clear that bnusA had relatively higher La-binding and relatively lower Yb-binding. Similarly, dSO 4685 had a similar level of La-binding compared to other in-range strains, but bound a much greater amount of Yb, implying relative increase in preference for heavy versus light REE.

[0089] Discussion of Examples

[0090] The described genetic screen for biosorption reveals layers of the outer surfaces (inner and outer membranes, and periplasmic layer) of S. oneidensis that modulate access to REE-binding sites in S. oneidensis. These layers include polysaccharides (synthesized by Polysaccharide Synthesis Operons 1 and 2), MSHA pili (synthesized and assembled by the MSHA Pilus Assembly Operon), and a variety of outer membrane proteins (SO_0456, SO 3099, MshQ, MshL, MshJ).

[0091] Disruption of Polysaccharide Synthesis Operon 1 Raises Biosorption [0092] Many of the genes coded by PSI are responsible for the synthesis of O- antigens, a major component of the lipopolysaccharide (LPS) layer on the outer membrane of S. oneidensis ⁶¹ . Disruption of all genes selected for further analysis in PSI increase total biosorption under all solution conditions (Figures 3A-H) and generally increase relative Yb- binding and decrease relative La-binding in high ionic strength conditions (Figure 4E). Among the three genes tested in the PSI group, only dSO 3183 is directly implicated in polysaccharide synthesis, although the similar biosorption effects of each of our three mutants suggest that they may all be part of a single pathway.

[0093] O-antigens themselves have not been directly implicated in REE-biosorption, but it has been theorized that phosphate groups on LPS components below the O-antigens, such as Lipid A, could be responsible for some REE-binding ³² . It is possible that knocking out these genes eliminates certain O-antigens, exposing these phosphate groups on the membrane. Phosphate groups tend to have a stronger affinity for heavier REE, and this could explain why there was a relative increase in biosorption of heavier REE.

[0094] Disruption of Polysaccharide Synthesis Operon 2 Modifies the Cell Membrane and REE Biosorption

[0095] The disruption mutants selected for in-depth analysis belonging to Polysaccharide Synthesis Operon 2 (PS2) generally cause significant increases in REE- biosorption in at least some cases, although the results were not necessarily consistent from gene to gene. For example, while wbpP and dwbpA significantly raise biosorption only in the low ionic strength cases, wzd increases biosorption in every case and wzz fails to significantly alter total biosorption at all. We suspect this is due to each of the genes in this group having a unique impact on biosorption.

[0096] It is considered that disruption of WbpP ( wbpP) raises REE-biosorption in low ionic strength conditions due to its significant role in membrane composition. WbpP transforms UDP-N-acetyl-D-glucosamine to the UDP-N-acetyl-D-galactosamine ⁶⁸ . In P. aeruginosa, WbpP plays a role in the synthesis of B-band O-antigens, a component of the lipopolysaccharide layer ⁶⁹ . In V. vulnificus, deletion of wbpP causes the failure of CPS (capsular polysaccharide) formation ⁷⁰ . The deletion also results in increased cell aggregation, hydrophobicity, and adherence to abiotic surfaces, all suggestive of substantial membrane changes ⁷⁰ . The exact mechanism for wbpP having increased biosorption in only the low ionic strength cases is not known. However, one possibility could be that B-band O-antigens are deleted. Consequently, binding sites that are normally covered by those O-antigens could be revealed that bind to REE only in low ionic strength environments.

[0097] Likewise, disruption of wbpA ( wbpA) also raises biosorption in low ionic strength conditions and plays a role in membrane composition. WbpA, like WbpP, takes UDP-N-acetyl-D-glucosamine as a substrate (but transforms it to uronic acid ⁶⁸ instead), and is thought to be a key protein in O-antigen biosynthesis in P. aeruginosa ¹¹ .

[0098] MSHA Genes Have Highly Environmentally Dependent Effects on Biosorption

[0099] The influence of the interaction of the solution environment with gene disruption on biosorption is illustrated by disruptions of the MSHA Pilus Assembly Operon (MSHA) genes. While 6/7 of the gene disruptions tested (all except mshC) had lower biosorption in the original screen, only 3/7 had significantly lower biosorption in any of the solution conditions selected for follow up testes and none of them had significantly lower biosorption for every solution condition.

[0100] MshA is responsible for forming the main subunit of the pilus and knocking it out thus has a major impact on the MSHA pilus ⁷² . mshA had significantly lower biosorption in the low ionic strength conditions. This seems to suggest that the MSHA pilus plays an important role in binding to REE in low ionic strength cases. At the same time, bmshA has no significant change in biosorption in the high ionic strength conditions, suggesting that the high NaCl concentration is preventing REE from binding to the pili.

[0101] Disruption of Pyrimidine Synthesis Group Increases REE-Biosorption Under High Ionic Strength, Low REE Conditions

[0102] Disruption mutants of the Pyrimidine Synthesis Group genes pyrC,pyrD, and pyrE all increase biosorption in the high ionic strength, low REE condition. These genes form a section of a pathway in pyrimidine metabolism that produces Orotidine 5 '-monophosphate (OMP) from carbamoyl aspartate.

[0103] Binding Site Changes from Single Gene Knockouts Tend to Have Multiple Effects

[0104] While some gene disruption mutants analyzed in this disclosure had consistent biosorption changes across every condition tested, for many other genes, changes in biosorption levels were inconsistent across different solution conditions. Some of these discrepancies have straightforward explanations. For example, several gene insertion mutants (such as polysaccharide synthesis protein wbpP) had higher biosorption for low ionic strength, but similar biosorption to the qWT for high ionic strength. In the case of wbpP, it is possible that this is because elimination of wbpP results in increased accessibility to binding sites that are capable of binding to REE in low ionic strength cases but not high ionic strength cases — possibly due to competition with sodium ions for binding sites. Thus, knocking out wbpP increases REE biosorption in the low ionic strength cases, but has no effect in the high ionic strength cases.

[0105] It could be expected that most gene insertion/deletions that affect biosorption would affect a single binding site. That the gene would encode a single outer membrane protein or a protein that alters some single REE binding lipid or polysaccharide found on the membrane. Based on the described genetic screen results, this appears to rarely be the case. More often, gene knockouts likely cause a cascade of effects on other genes resulting in changes to multiple binding sites. Gene disruptions can also have confounding effects unrelated to REE binding site composition. It is possible that some gene disruptions impact the shape of the bacteria or perhaps the optical density to bacteria ratio (since the optical density is what is used to normalize the bacterial density from assay to assay). Alternatively, a mutation could alter the OD: surface area ratio, making it an artificial outlier in the REE- biosorption screen without truly affecting biosorption. It is known that growth phase can impact metal biosorption ⁷³ . A gene knockout might not affect membrane composition but instead introduce a growth defect that affects the bacterial growth phase, and hence biosorption, at the sampled optical density.

[0106] Changes to Lanthanide Preference Improves REE Purification Process

[0107] To analyze the size of the lanthanide preference changes created by the mutants we isolated, we created a model to get an order of magnitude estimate for how these preference changes affect a lanthanide purification process (Figures 11 and 2A-F, Table S3, Note SI). In Figure 11 and Note SI the disclosure provides a simplified system for lanthanide enrichment that uses repeated biosorption and elution. A mixed solution of REE is run through a column and allowed to bind to immobilized bacteria. After equilibration the free fraction (the liquid) is removed from the column and moved to a wash collection container. Next, the bound fraction is eluted (for example by a pH swing ^2,74 ). The eluant is reloaded (or sent to another column) for further purification. In this model, we follow the biosorbed/eluted REE solution at each stage of the process. The system is designed such that 50% of REE are biosorbed to the bacteria at each stage (Note SI). We also assume that there is only a single binding site type on the bacteria (Note SI).

[0108] We compare the enrichment process for one strain, 5wbpA, and our calculated baseline (Figure 12 and Table S3). The baseline is what an average transposon insertion strain’s REE biosorption distribution would look like if it biosorbed the same total amount of REE as dwbpA (see Figure 4E for details about this baseline). We note that, since there are apparently multiple binding sites on each strain, we cannot definitively determine what the separation factor between the REE will be as the REE distribution changes. We chose to look at dwbpA (low ionic strength) as a representative embodiment, as it has similar separation factor improvements in environments with different starting REE concentrations.

[0109] We calculated Eu enrichment from an equimolar solution of Eu, Yb and La by <ywbpA, using measured separation factors from the low ionic strength solution conditions (LL and LH). Despite the two conditions (LH and LL) having different percent increases in europium biosorption (1% and 2% respectively), we obtained similar results. For the LL condition, we found dwbpA produced a 27% reduction in the number of steps (30 for the baseline, 22 steps for the mutant) compared to the baseline required to produce a 99% purity europium solution. In the LH condition, SwbpA produced a 25% reduction (28 for the baseline, 21 for the mutant) in the number of steps to 99% purity (Table S3).

[0110] The genes and gene clusters described herein are shown to be involved in Eu- biosorption and directly involved in membrane composition.

[0111] Many of the gene disruptions that affect total REE-binding and modify the REE-binding preference of S. oneidensis likely have functions that directly affect the outer surface. Both Polysaccharide Synthesis Operon 1 and 2 are involved in making O-antigens on the lipopolysaccharide layer. The MSHA genes synthesize pili on the outer membrane of S. oneidensis. Two members of the Miscellaneous Group are related to other outer membrane structures: SO 2183 is involved in synthesis of the peptidoglycan layer and the SO 4685 protein is involved in synthesis of capsular polysaccharides.

[0112] Two individual gene disruptions stood out in their impact on total biosorption. In every solution condition, disrupting SO 4685 resulted in the highest biosorption observed, while disrupting nusA resulted in the lowest biosorption observed.

[0113] The disclosure demonstrates 10 gene disruptions or deletions affect the biosorption selectivity preference. Despite the apparently small size of changes to biosorption preference (~ 1 to 4%; Figure 4) caused by disruptions to single genomic loci, these changes might produce large reductions in the length of a repeated enrichment process for individual lanthanides (Figure 12). As a non-limiting example, in a simplified model of S. oneidensis binding (Figure 11, the 2% increase in europium binding for SwbpA resulted in up to a 27% reduction in the number of enrichment steps needed to reach 99 and 99.9% purity (Figure 12, Note SI, Table S3). The disclosure includes using a knockout of wbpA plus up-regulation of pyrD produced by swapping its endogenous promoter for a stronger version (e.g., J23100 or Tet promoter ⁷⁵ ) to produce a mutant with even better separation characteristics.

[0114] The disclosure includes use of multi-site genome engineering tools (e.g., MAGE ⁴⁷ ) to mutate regulatory and coding regions of genes identified in this disclosure to alter the selectivity of S. oneidensis" membrane for REE. Iterative applications of this process are included to create S. oneidensis strains with increased affinity for specific REE. In a nonlimiting embodiment, a very good mutant may have 10-50 genes that control the membrane composition downregulated or deleted (including possibly wbpA) and 10-50 or more genes upregulated (possibly including pyrD). Targeted multiplexed mutagenesis could produce a mutant like this with a few rounds of mutagenesis. [0115] Bibliography

1 Thiele, N. A., Fiszbein, D. J., Woods, J. J. & Wilson, J. J. Tuning the Separation of Light Lanthanides Using a Reverse-Size Selective Aqueous Complexant. Inorg Chem 59, 16522-16530 (2020). doi.org:10.1021/acs.inorgchem.0c02413

2 Bonificio, W. D. & Clarke, D. R. Rare-Earth Separation Using Bacteria. Environmental Science & Technology Letters 3, 180-184 (2016). doi.org: 10.1021/acs.estlett.6b00064

3 Baym, M., Shaket, L., Anzai, I. A., Adesina, O. & Barstow, B. Rapid construction of a whole-genome transposon insertion collection io Shewanella oneidensis\yy Knockout Sudoku. Nature Communications 7, 13270 (2016). doi.org: 10.1038/ncommsl3270

4 Anzai, I. A., Shaket, L., Adesina, O., Baym, M. & Barstow, B. Rapid curation of gene disruption collections using Knockout Sudoku. Nature Protocols 12, 2110-2137 (2017). doi.org: 10.1038/nprot.2017.073

5 Dent, P. C. Rare earth elements and permanent magnets. Journal of Applied Physics 111, 07A721 (2012). doi.org: 10.1063/1.3676616

6 Lucas, J., Lucas, P., Le Mercier, T., Rollat, A. & Davenport, W. Rare Earths: Science, Technology, Production and Use. (Elsevier Inc., 2014).

7 Schubert, E. F. & Kim, J. K. Solid-State Light Sources Getting Smart. Science 308, 1274-1278 (2005). doi.org: 10.1126/science.1108712

8 Norman, A. F., Prangnell, P. B. & McEwen, R. S. The solidification behaviour of dilute aluminium-scandium alloys. Acta Mater 46, 5715-5732 (1998). doi.org: 10.1016/sl359-6454(98)00257-2

9 Adesina, O., Anzai, I. A., Avalos, J. L. & Barstow, B. Embracing Biological Solutions to the Sustainable Energy Challenge. Chem 2, 20-51 (2017). doi.org: 10.1016/j.chempr.2016.12.009

10 Nazarov, M. & Noh, D. Y. New generation of europium and terbium activated phosphors : from syntheses to applications. (Pan Stanford Publishing, 2011).

11 Muller, T. & Friedrich, B. Development of a recycling process for nickel-metal hydride batteries. J Power Sources 158, 1498-1509 (2006). doi.org: 10.1016/j.jpowsour.2005.10.046

12 International Energy Outlook 2021 with Projections to 2050. (2021).

13 Annual Energy Outlook 2020, with projections to 2050. (U.S. Energy Information Administration, Office of Energy Analysis, U.S. Department of Energy, Washington, DC, 2020).

14 Voncken, J. H. L. The Rare Earth Elements, An Introduction. (2016). 15 Scheyder, E. China Set to Control Rare Earth Supply for Years Due to Processing Dominance, <www.reuters.com/article/us-china-usa-rareearth-refining/ china-set-to- control-rare-earth-supply-for-years-due-to-processing-domina nce-idUSKCNlT004J> (2019).

16 Mining the Future: How China is set to dominate the next Industrial Revolution. (Foreign Policy, 2019).

17 Peiravi, M. el al. Chemical extraction of rare earth elements from coal ash. Minerals & Metallurgical Processing 34, 170-177 (2017). doi.org:10.19150/mmp.7856

18 in CRC Handbook of Chemistry and Physics, 102nd Edition (Taylor andFrancis, 2021).

19 Peiravi, M. el al. A Review of Rare-Earth Elements Extraction with Emphasis on Non- conventional Sources: Coal and Coal Byproducts, Iron Ore Tailings, Apatite, and Phosphate Byproducts. Mining, Metallurgy & Exploration 38, 1-26 (2021). doi.org: 10.1007/s42461 -020-00307-5 0 Kuan, S. H., Saw, L. H. & Ghorbani, Y. in International Annual Symposium on Sustainability Science and Management 105. 1 Johnson, D. B. Biomining — biotechnologies for extracting and recovering metals from ores and waste materials. Current Opinion in Biotechnology 30, 24-31 (2014). doi.org: 10.1016/j. copbio.2014.04.008 2 Reed, D. W., Fujita, Y., Daubaras, D. L., Jiao, Y. & Thompson, V. S. Bioleaching of rare earth elements from waste phosphors and cracking catalysts. Hydrometallurgy 166, 34-40 (2016). doi.org: 10.1016/j .hydromet.2016.08.006 3 Schmitz, A. M. el al. Generation of a Gluconobacter oxydans knockout collection for improved extraction of rare earth elements. Nature Communications 12, 6693 (2021). doi.org: 10.1038/s41467-021 -27047-4 4 Cotruvo, J. A., Featherston, E. R., Mattocks, J. A., Ho, J. V. & Laremore, T. N. Lanmodulin: A Highly Selective Lanthanide-Binding Protein from a Lanthanide- Utilizing Bacterium. Journal of the American Chemical Society, 15056-15061 (2018). doi.org: 10.1021/jacs.8b09842 5 Deblonde, G. J. el al. Selective and Efficient Biomacromolecular Extraction of Rare- Earth Elements using Lanmodulin. Inorg Chem 59, 11855-11867 (2020). doi . org : 10.1021 / acs. inorgchem. OcO 1303 6 Park, D. M. et al. Bioadsorption of Rare Earth Elements through Cell Surface Display of Lanthanide Binding Tags. Environmental Science and Technology 50, 2735-2742 (2016). doi . org : 10.1021 /acs. est.5b06129 27 Tay, P. K. R., Manjula-Basavanna, A. & Joshi, N. S. Repurposing bacterial extracellular matrix for selective and differential abstraction of rare earth elements. Green Chemistry 20, 3512-3520 (2018). doi.org: 10.1039/c8gc01355a

28 Bogart, J. A., Lippincott, C. A., Carroll, P. J. & Schelter, E. J. An Operationally Simple

Method for Separating the Rare - Earth Elements Neodymium and Dysprosium. Angewandte Chemie International Edition 54, 8222-8225 (2015). doi.org: 10.1002/anie.201501659

29 Higgins, R. F. el al. Magnetic Field Directed Rare - Earth Separations. Angewandte Chemie International Edition 59, 1851-1856 (2020). doi.org: 10.1002/anie.201911606

30 Lin, W. et al. Promising priority separation of europium from lanthanide by novel DGA-functionalized metal organic frameworks. Minerals Engineering 164, 106831 (2021 ). doi . org: 10.1016/j .mineng.2021.106831

31 Yang, H. et al. Selective Crystallization of Rare - Earth Ions into Cationic Metal -

Organic Frameworks for Rare - Earth Separation. Angewandte Chemie International Edition 60, 11148-11152 (2021). doi.org: 10.1002/anie.202017042

32 Park, D. M., Brewer, A., Reed, D. W ., Lammers, L. N. & Jiao, Y. Recovery of Rare

Earth Elements from Low-Grade Feedstock Leachates Using Engineered Bacteria. Environmental Science and Technology 51, 13471-13480 (2017). doi.org: 10.1021/acs.est.7b02414

33 Brewer, A. et al. Recovery of Rare Earth Elements from Geothermal Fluids through Bacterial Cell Surface Adsorption. Environmental Science and Technology 53, 7714- 7723 (2019). doi.org: 10.1021/acs.est.9b00301

34 Brewer, A. et al. Microbe Encapsulation for Selective Rare-Earth Recovery from Electronic Waste Leachates. Environmental Science and Technology 53, 13888-13897 (2019). doi.org: 10.1021/acs.est.9b04608

35 Good, N. M. et al. Hyperaccumulation of Gadolinium by Methylorubrum extorquens AMI Reveals Impacts of Lanthanides on Cellular Processes Beyond Methyl otrophy. Front Microbiol 13, 820327 (2022). doi.org: 10.3389/fmicb.2022.820327

36 Hu, A., MacMillan, S. N. & Wilson, J. J. Macrocyclic Ligands with an Unprecedented Size-Selectivity Pattern for the Lanthanide Ions. J Am Chem Soc 142, 13500-13506 (2020). doi . org: 10.1021/j acs.0c05217 Abbas, S. H., Ismail, I. M., Mostafa, T. M. & Sulaymon, A. H. Biosorption of heavy metals: a review. J Chem Set Technol 3, 74-102 (2014). Microbial Biosorption of Metals. (Springer Dordrecht, 2011). Tang, X. et al. Profiling the Membrane Proteome of Shewanella oneidensis MR-1 with New Affinity Labeling Probes. Journal of Proteome Research 6, 724-734 (2007). doi.org: 10.1021/pr060480e PMID - 17269728 Sohlenkamp, C. & Geiger, O. Bacterial membrane lipids: diversity in structures and pathways. FEMS Microbiol Rev 40, 133-159 (2016). doi.org: 10.1093/femsre/fuv008 Perez-Burgos, M. & Sogaard- Andersen, L. Biosynthesis and function of cell-surface polysaccharides in the social bacterium Myxococcus xanthus. Biol Chem 401, 1375- 1387 (2020). doi.org: 10.1515/hsz-2020-0217 Takahashi, Y., Chatellier, X., Hattori, K. H., Kato, K. & Fortin, D. Adsorption of rare earth elements onto bacterial cell walls and its implication for REE sorption onto natural microbial mats. Chemical Geology 219, 53-67 (2005). doi . org : 10.1016/j . chemgeo.2005.02.009 Fomina, M. & Gadd, G. M. Biosorption: current perspectives on concept, definition and application. Bioresource technology 160, 3-14 (2014). doi. org: 10.1016/j.biortech.2013.12.102 Moriwaki, H. et al. Application of Freeze-Dried Powders of Genetically Engineered Microbial Strains as Adsorbents for Rare Earth Metal Ions. ACS Applied Materials and Interfaces 8, 26524-26531 (2016). doi.org: 10.1021/acsami.6b08369 Takahashi, Y., Yamamoto, M., Yamamoto, Y. & Tanaka, K. EXAFS study on the cause of enrichment of heavy REEs on bacterial cell surfaces. Geochimica et Cosmochimica Acta 7 , 5443-5462 (2010). doi.org: 10.1016/j.gca.2010.07.001 Moriwaki, H., Koide, R., Yoshikawa, R., Warabino, Y. & Yamamoto, H. Adsorption of rare earth ions onto the cell walls of wild-type and lipoteichoic acid-defective strains of Bacillus subtilis. Applied Microbiology and Biotechnology 97, 3721-3728 (2013). doi.org: 10.1007/s00253-012-4200-3 PMID - 22684329 Wang, H. H. etal. Programming cells by multiplex genome engineering and accelerated evolution. Nature 460, 894 898 (2009). doi.org: 10.1038/nature08187 Raman, S., Rogers, J. K., Taylor, N. D. & Church, G. M. Evolution-guided optimization of biosynthetic pathways. Proceedings of the National Academy of Sciences 111, 17803-17808 (2014). doi.org: 10.1073/pnas. l409523111 49 Halperin, S. O. et al. CRISPR-guided DNA polymerases enable diversification of all nucleotides in a tunable window. Nature 560, 248—252 (2018). doi.org: 10.1038/s41586-018-0384-8

50 Corts, A. D. Efficient andPrecise Genome Editing in Shewanella with Recombineering and CRISPR/Cas9-mediated Counter-selection PhD thesis, University of Minnesota, (2019).

51 Corts, A. D., Thomason, L. C., Gill, R. T. & Gralnick, J. A. Efficient and Precise

Genome Editing in Shewanella with Recombineering and CRISPR/Cas9-Mediated Counter-Selection. ACS Synth Biol 8, 1877-1889 (2019). doi.org: 10.102 l/acssynbio.9b00188

52 Corts, A. D., Thomason, L. C., Gill, R. T. & Gralnick, J. A. A new recombineering system for precise genome-editing in Shewanella oneidensis strain MR-1 using singlestranded oligonucleotides. Sci Rep 9, 39 (2019). doi.org: 10.1038/s41598-018-37025-4

53 Baym, M., Shaket, L., Anzai, I. A., Adesina, O. & Barstow, B. Rapid construction of a whole-genome transposon insertion collection for Shewanella oneidensis by Knockout Sudoku. Nature Communications 7, 13270 (2016). doi.org: 10.1038/ncommsl3270

54 Hogendoorn, C. et al. Facile Arsenazo Ill-Based Assay for Monitoring Rare Earth

Element Depletion from Cultivation Media for Methanotrophic and Methyl otrophic Bacteria. AppL Environ. Microb. 84, e02887-02817 (2018). doi.org: 10.1128/aem.02887-17

55 Ashbumer, M. et al. Gene ontology: tool for the unification of biology. Nature genetics 25, 25-29 (2000).

56 Consortium, G. O. The Gene Ontology resource: enriching a GOld mine. Nucleic Acids Res 49, D325-D334 (2021). doi.org: 10.1093/nar/gkaal 113

57 Dehal, P. S. et al. MicrobesOnline: an integrated portal for comparative and functional genomics. Nucleic Acids Res 38, D396-400 (2010). doi.org: 10.1093/nar/gkp919

58 Taboada, B., Ciria, R., Martinez-Guerrero, C. E. & Merino, E. ProOpDB: Prokaryotic

Operon DataBase. Nucleic Acids Res 40, D627-631 (2012). doi.org: 10.1093/nar/gkrl020

59 Lassak, J., Bubendorfer, S. & Thormann, K. M. Domain analysis of ArcS, the hybrid sensor kinase of the Shewanella oneidensis MR-1 Arc two-component system, reveals functional differentiation of its two receiver domains. JBacteriol 195, 482-492 (2013). doi.org: 10.1128/JB.01715-12 60 luchi, S. & Lin, E. arcA (dye), a global regulatory gene in Escherichia coli mediating repression of enzymes in aerobic pathways. Proceedings of the National Academy of Sciences 85, 1888-1892 (1988).

61 Heidelberg, J. F. et al. Genome sequence of the dissimilatory metal ion-reducing bacterium Shewanella oneidensis. Nature Biotechnology 20, nbt749 (2002). doi.org: 10.1038/nbt749

62 Rowe, A. R. et al. Identification of a Pathway for Electron Uptake in Shewanella oneidensis. Communications Biology 4, 957 (2021). doi.org: 10.1038/s42003-021- 02454-x

63 Fitzgerald, L. A. et al. Shewanella oneidensis MR-1 Msh pilin proteins are involved in extracellular electron transfer in microbial fuel cells. Process Biochemistry 47, 170-174 (2012). doi.org: 10.1016/j.procbio.2011.10.029

64 Naville, M., Ghuillot-Gaudeffroy, A., Marchais, A. & Gautheret, D. ARNold: a web tool for the prediction of Rho-independent transcription terminators. RNA Biol 8, 1 1 - 13 (2011). doi.org: 10.4161/rna.8.1.13346

65 Jacobs, M. A. et al. Comprehensive transposon mutant library of Pseudomonas aeruginosa. Proceedings of the National Academy of Sciences 100, 14339-14344 (2003). doi.org: 10.1073/pnas.2036282100

66 Opijnen, T. v., Bodi, K. L. & Camilli, A. Tn-seq: high-throughput parallel sequencing for fitness and genetic interaction studies in microorganisms. Nat Methods 6, 767-772 (2009). doi . org : 10.1038/nmeth.1377

67 Bertani, B. & Ruiz, N. Function and Biogenesis of Lipopolysaccharides. EcoSal Plus 8 (2018). doi . org: 10.1128/ecosalplus.ESP-0001-2018

68 Edel, M. et al. Extracellular riboflavin induces anaerobic biofilm formation in Shewanella oneidensis. Biotechnol Biofuels 14, 130 (2021). doi.org: 10.1186/sl3068- 021-01981-3

69 Belanger, M., Burrows, L. L. & Lam, J. S. Functional analysis of genes responsible for the synthesis of the B-band O antigen of Pseudomonas aeruginosa serotype 06 lipopolysaccharide. Microbiology+ 145, 3505-3521 (1999). doi.org: 10.1099/00221287-145-12-3505

70 Lee, K. J., Kim, J. A., Hwang, W ., Park, S. J. & Lee, K. H. Role of capsular polysaccharide (CPS) in biofilm formation and regulation of CPS production by quorum-sensing in Vibrio vulnificus. Mol Microbiol 90, 841-857 (2013). doi.org: 10. I l l 1/mmi.12401 71 Miller, W. L. et al. Biochemical characterization of WbpA, a UDP-N-acetyl-D- glucosamine 6-dehydrogenase involved in O-antigen biosynthesis in Pseudomonas aeruginosa PAO1. J Biol Chem 279, 37551-37558 (2004). doi.org: 10.1074/jbc.M404749200

72 Thormann, K. M., Saville, R. M., Shukla, S., Pelletier, D. A. & Spormann, A. M. Initial Phases of biofilm formation in Shewanella oneidensis MR-1. J Bacteriol 186, 8096- 8104 (2004). doi.org: 10.1128/JB.186.23.8096-8104.2004

73 Daughney, C. J., Fowle, D. A. & Fortin, D. The effect of growth phase on proton and metal adsorption by Bacillus subtilis. Geochimica et Cosmochimica Acta 65, 1025- 1035 (2001). doi.org: 10.1016/s0016-7037(00)00587-l

74 Park, D. et al. A biosorption-based approach for selective extraction of rare earth elements from coal byproducts. Separation and Purification Technology 241, 116726 (2020). doi . org: 10.1016/j . seppur.2020.116726

75 Yi, Y.-C. & Ng, I. S. Establishment of toolkit and T7RNA polymerase/promoter system in Shewanella oneidensis MR- 1. Journal of the Taiwan Institute of Chemical Engineers 109, 8-14 (2020). doi.org: 10.1016/j.jtice.2020.02.003

76 Yamamoto, N. et al. Update on the Keio collection of Escherichia coli single-gene deletion mutants. Mol SystBiol 5, 335 (2009). doi.org: 10.1038/msb.2009.92

77 Rohwer, H., Collier, N. & Hosten, E. Spectrophotometric study of arsenazo III and its interactions with lanthanides. Anal Chim Acta 314, 219-223 (1995). doi.org: 10.1016/0003-2670(95)00279-9

78 Buchfink, B., Reuter, K. & Drost, H.-G. Sensitive protein alignments at tree-of-life scale using DIAMOND. Nat Methods 18, 366-368 (2021). doi.org:10.1038/s41592- 021-01101-x

79 Gotz, S. et al. High-throughput functional annotation and data mining with the

Blast2GO suite. Nucleic Acids Research 36, 3420-3435 (2008). doi.org: 10.1093/nar/gknl76

80 Adrian, A. & Jorg, R. Gene Set Enrichment Analysis with topGO, <http://compdiag.molgen.mpg.de/ngfn/docs/2007/sep/topGO_E xercises.pdf (2007). [0116] Materials and Methods

[0117] Media Preparation

[0118] bm20 Media

[0119] bm20 is composed of 20 mM MES buffer, 8.6 mM ammonium chloride, 0.5 mM magnesium sulfate, 1.7 mM ammonium sulfate, and 5 mL/L Trace Mineral Supplement (ATCC). The media was adjusted to pH 5.5 with 1 M NaOH.

[0120] Genome-wide REE Biosorption Screen

[0121] Introduction

[0122] We screened the S. oneidensis whole genome knockout collection ^4,53 for biosorption of the rare earth element (REE) europium. We hypothesized that using europium, which lies in the middle of the REE size range, would allow us to maximize the number of genes we could discover.

[0123] We screened a subset of the knockout collection comprising 67 96-well plates that covers 3,472 unique genes (the S. oneidensis knockout collection is comprised of ~ 50% blank wells). Due to some mutants in the collection failing to grow, we screened 3,373 unique genes in total. Cross-contamination did not appear to be a problem as, on average, we had only 1.6 contaminated wells (wells that were not supposed to have bacteria in them yet grew anyway) per plate. The maximum number of contaminated wells in a plate was 10. If we assume that the filled-to-filled cross-contamination rate is the same as the filled-to-blank rate, then we get a total contamination rate of 1.6/48 = 3.3%. To put this into context, the mis-labeling rate on the original gold standard Keio whole genome knockout collection for E. coli was 4% ⁷⁶ . Second, given that most mutants (3,373 - 242 = 3,131 of them) have a similar biosorption phenotype, cross-contamination is far more likely to mask mutants with an outlier biosorption phenotype than create an artificial hit.

[0124] The genetic screen was conducted over the course of several weeks and was divided into batches, each of which took two days to process. Typical batch sizes were between 4 to 8 plates.

[0125] Replication of Shewanella oneidensis Whole-genome Knockout Collection [0126] Microplates from the S. oneidensis knockout collection were replicated from a master collection (stored at -80 °C) with a pin-tool replicator (Enzy Screen CR1000) into a flat-bottom polypropylene plate (Part no. 655261, Greiner) containing 150 pL of LB media per well with 30 mg L' ¹ Kanamycin. Newly inoculated plates were incubated at 30 °C shaking at 800 rpm in a high-throughput microplate shaker (Infers Multitron Pro) for between 16 and 20 hours. The following morning, 3 pL of culture was transferred to a new plate containing fresh LB with 30 mg L' ¹ Kanamycin. The newly diluted cultures were grown for 5-6 hours.

[0127] Biosorption Assay

[0128] Upon removing our plates from the incubator, we diluted 40 pL of culture from each plate into a new round-bottom polystyrene 96-well plate (Part no. 650101, Greiner) with 150 pL of bm20 media (see Media Preparation). After this transfer, we took an optical density (OD) measurement of our plate with a plate reader (Biotek Synergy 2) at 590 nm. We called this OD the “growth OD” because it was proportional to the final OD that the bacteria grew to.

[0129] After taking the OD, we centrifuged each plate in a swinging bucket centrifuge with micro-well plate adaptors (Eppendorf 581 OR) at a speed of 3,214 * g for 7 minutes. After that, we positioned a 96-well pipette to remove supernatant from the edges of the wells of the plate. We removed as much supernatant as possible because components of the growth media (data not shown) can bind to rare earth elements and interfere with biosorption. We then rinsed the bacteria by adding 170 pL of bm20 to each well and resuspended by shaking. We then repeated the centrifuging and removing of supernatant steps. After these two rinsing steps, the concentration of growth media components, including the kanamycin, should have been negligible.

[0130] We added 200 pL of bm20 with approximately 25 pM of europium to the rinsed S. oneidensis cells and resuspended by vortexing. We then shook the plate in the plate reader or in a plate vortexer for 10 minutes while biosorption occurred. Previous research has shown that bacterial cells tend to reach full biosorption capacity within 10 minutes ^33,42 . The shaking also helps to prevent bacterial aggregation. Upon the completion of the shaking step, we took another OD measurement (called the “final OD”) to have an estimate of how many bacteria were present in each well for the biosorption assay.

[0131] We centrifuged each plate once more at a speed of 3214 * g for 10 minutes. We then transferred 100 pL of the supernatant to a new flat bottom polystyrene plate (Greiner bio-one ref: 655101). We used the plate reader to take a spectrum of absorbance from 580 nm to 680 nm in increments of 10 nms. These are our ’’blank” measurements. We then added

100 pL of a solution of 60 pM concentration of Arsenazo III dye dissolved in pH 3.5 20 mM MES buffer to each well with the 96-well pipette. Arsenazo III has been established to be a reliable indicator for free REE concentration at this pH ^54,77 . We shook the plate for 4 minutes and then we used the plate reader to measure the absorbance in the same wavelength range as the blanks. These absorbance measurements served as a proxy measurement for the concentration of rare earth elements that were not adsorbed by the bacteria.

[0132] Analysis of Genome-wide REE Biosorption Screen

[0133] Challenges of Identification of Mutants with Differential Biosorption

[0134] Optical density and As-III absorbance measurements were corrected to account for blemishes on the exterior of assay plates, especially due to centrifugation. We thus took our final optical density measurement to be the minimum of the optical density prior to rinsing and the optical density after we have finished preparing our biosorption assay. Likewise, As-III absorbance measurements were background corrected by subtracting a blank measurement taken prior to As-III addition.

[0135] Non-uniform growth is the biggest challenge in identification of mutants with genetic modifications that affect REE biosorption. Due to the dynamics of ligand-receptor binding chemistry, we would expect that cultures with higher densities will produce higher overall biosorption but have lower biosorption per cell.

[0136] We used deviations from a linear piecewise relationship between optical density at 590 nm (OD590) and As-III absorbance to identify mutants with truly differential biosorption (Figure IB). The OD to As-III absorbance function varied from plate to plate, likely due to slightly different growth conditions. As a result, we treated every plate separately when identifying outliers.

[0137] The linear piecewise function relating OD590 and As-III is made of 2 sections. For OD590 ~ 0.2, As-III absorbance was linearly related to OD590. For OD590 ~ 0.2, we took advantage of the large number of blank wells in each 96-well plate to measure As-III absorbance with no biosorption. Thus, for the earlier part of the piecewise function, we fit a line between a point marking the average OD590 and As-III absorbance of blank wells and the point with the smallest OD590 in the set used to create the second part of the function. We found that most points in each plate generally fell along our piecewise function.

[0138] To find biosorption outliers, we calculated the standard deviation of the distances of each point from the linear piecewise function for each plate. We marked any point as significant if it had a distance of more than 2 standard deviations from our piecewise function (Figure IB).

[0139] Arsenazo III Assay Quality Control

[0140] Quality-control procedures were used to identify low-quality data during analysis. Our procedures specifically looked for anomalous optical density measurements and absorbance spectra patterns. [0141] Plots of final optical density versus growth optical density form a compact line (Figure 5A). Datapoints where the final OD590 was more than three standard deviations above the average value were flagged. Datapoints where the final OD590 was below average were not flagged, as these indicated either that the growth OD590 measurement was off (which would not affect the final analysis) or that there was a large loss of bacteria during rinsing (which also would not affect the final analysis).

[0142] Arsenazo III absorbance measurements were quality-controlled by ratiometric analysis. Errors in the As-III absorbance measurement can arise due to errors in pipetting the As-III stock, or by the presence of dust or scratches on the assay plate surfaces. For the range of As-III absorbance measurements observed in our assay, the relationship between the 650 nm absorbance and the 680:650 absorbance ratio was linear (Figure 5). Extreme outliers could occasionally skew this analysis, so we eliminated the five data points furthest from the original linear fit and created a second line of best fit. Data points that lay more than three standard deviations away from the line of best fit were flagged.

[0143] Manual inspection was used as the final step in quality control resulting in the selection of 240 mutants for further analysis. Each of the 294 data points that had outlier biosorption measurements in the screen were manually examined. We paid special attention to the datapoints flagged with a high final OD590 or with an anomalous absorbance spectrum. We eliminated 12 datapoints that were listed as blank in the S. oneidensis knockout collection catalog, but that displayed cross-contamination. We eliminated 7 datapoints that lacked transposon location information. Finally, we removed strains where we judged that the OD590 to As-III absorbance piecewise function did not provide a reliable estimate for significance due to lack of surrounding data.

[0144] Gene Ontology Enrichment Analysis

[0145] We followed the gene ontology enrichment analysis procedures laid out in Schmitz et al ² , except we did not use InterProScan to collect gene ontology data. In brief, we used DIAMOND ⁷⁸ to assign annotated protein models with the closest BLAST hit using the Uniref90 database (downloaded from www.uniprot.org/uniref/), an E-value threshold of 10' ¹⁰ , and a block size of 10. We used the output of this search to assign gene ontologies with BLAST2GO ⁷⁹ .

[0146] We performed a gene ontology enrichment analysis with the BioConductor topGO package ⁸⁰ using the default weight algorithm, the TopGO Fisher test, with a p-value threshold of 0.05. [0147] We performed separate gene ontology enrichment analyses for mutants with significantly higher and lower biosorption (Figure 1C). Following the transposon mutant collection screen, we found that SSO 0625 and SglnA both produced lower Eu-biosorption and added them to the lower biosorption category.

[0148] Operon Enrichment Analysis

[0149] Operon enrichment analysis was used as a complement to ontology enrichment analysis to identify groups of genes involved in REE-biosorption (Figures 2A- 2F and Figure 8).

[0150] Operon memberships in the S. oneidensis genome were predicted by the union of results from operon predictions by MicrobesOnline ⁵⁷ and ProOpDB ⁵⁸ . In most cases, the two sources produced highly similar results.

[0151] We used Fisher’s exact test to calculate if the set of genes whose disruption conferred differential biosorption was enriched in operons with more than one hit. We compared the total number of gene disruptions that significantly affected biosorption (out of the total number of genes assayed in our genetic screen) to the number of gene disruptions within the operon that significantly affected biosorption (out of the total number of genes we looked at within that operon in our genetic screen). Fisher’s exact test was conducted with the fishertest function in MATLAB.

[0152] We also applied Fisher’s exact test to calculate if the sets of gene disruptions that either increased or decreased biosorption were enriched in operons that had more than one gene disruption that increased or decreased biosorption, respectively.

[0153] Spot Check of Mutants Identified in Whole Genome Screen

[0154] We sought to test how well our screen did at correctly identifying biosorption outliers. To this end, we re-screened a subset of both mutants we identified as outliers in our original screen and ones we identified as non-outliers. We chose this subset from plates seven through eleven of the collection. We analyzed all 32 outlier strains that appeared in these plates and 32 random non-outlier strains. We checkerboarded the strains (one blank well between every strain) and had three replicates for each strain. The replicates were spread out onto different areas of the plates.

[0155] The assay itself was carried out almost identically to the original screen. The only difference is that after replicating plates seven through eleven, we re-arrayed mutants of interest onto new plates. This was done using a Norgren Colony Picker. Besides this extra growth step, no other changes were made. We note that none of the blank wells had contamination in the final assay.

[0156] Analysis of the data was also very similar to the original screen. Instead of fitting a line of best fit to all the mutant data, we only fit the line of best fit to the non-outlier points. We also slightly modified our method of looking for absorbance spectrum outliers. We removed the manual checks of the spectrum to use uniformly applied criteria. We chose to instead look for absorbance spectra outliers by comparing the 650 nm wavelength to the ratio of the 660 nm and 650 nm. We did this because, in our manual absorbance analysis for the original screen, many of the ones that looked the most off had a noticeably irregular 660/650nm ratio.

[0157] Because we had three replicates instead of one, we also modified our method of determining which mutants were significant. In the original screen, we required a significance of 2 standard deviations from the mean. In our new screen, we required that at least two replicates had 1 standard deviation from the mean in the same direction. This may seem like a less stringent criteria since the expectation would be a normal distribution of data to have non-outlier points meet this criterion 17% of the time. Our points do not, however, follow a normal distribution. We believe that we occasionally have noisy points that contribute to the standard deviation but there are typically not more than one high-noise points per strain. Requiring only two points allows analysis of cases where a strain has a single high-noise point. The use of 1 standard deviation instead of 2 is also validated by the very small number of non-outliers in the original screen appearing to be outliers in this spot check.

[0158] Confirmation of Transposon Mutant Identity

[0159] We validated the identity of transposon mutants from the S. oneidensis knockout collection that we conducted follow up analyses on using site specific PCR. The verification reactions used a common primer that bound to the Himar transposon, and a mutant specific primer that bound to the genomic region predicted to be outside the transposon. The identities of all but one transposon mutant was correctly predicted in the S. oneidensis knockout collection catalog. The single mutant that was mis-identified was originally annotated as arcA (fiSO 3988), but later found to be 147 bp upstream of SO 2183.

[0160] Note on glnA and SO 0625

[0161] The 67 plates that we screened and conducted our analyses on contains mutants whose transposon locations are well characterized by double sequencing verification. This is validated by only a single mutant out of 25 tested having the insertion in an unexpected location. The collection, however, has 10 additional plates with mutants whose transposon locations had some ambiguity. We screened these plates and identified one mutant that we found to be a particularly large outlier in biosorption. The location for the transposon in this mutant was determined be in either glnA or SO 0625. We decided to make clean deletions for these two genes instead of isolating the transposon mutants.

[0162] Construction of Gene Deletion Mutants

[0163] Clean deletion mutants were constructed to validate the results of transposon screening for hplA. SO 4685, mshJ, and SO 3385 as well as for glnA and SO 0625.

Deletions were made by homologous recombination using a suicide vector containing a kanamycin resistance cassette flanked by 1000 bp upstream and downstream sequences surrounding the gene of interest. Mutants that had undergone a second recombination (removing the gene of interest and the kanamycin cassette) or reversion (where the gene of interest was recovered) were selected by a sucrose counter selection. Mutants with a clean deletion were separated from revertants by PCR screening ⁶² .

[0164] To ensure that the gene deletion process did not introduce additional changes to the S. oneidensis genome, we checked REE-biosorption by revertants recovered in the process of deleting two of the genes. In both cases, REE-biosorption was statistically indistinguishable from the true wild-type ( -value < 0.05) (Figure 9).

[0165] Analytical Measurement of Biosorption with ICP-MS

[0166] We explored biosorption in four different solution conditions (detailed in Table 1) with three rare earth elements: La (representing light REE), Eu (representing middle REE), and Yb (representing heavy REE). In every condition, bacterial culture density was normalized to the same optical density.

[0167] Bacterial strains of interest were retrieved from glycerol stocks frozen at -80 °C and recovered on LB agar plates (with 50 mg L' ¹ kanamycin for transposon insertion strains). We picked three single colonies for each strain and inoculated them into separate wells containing 200 pL of LB (with 50 mg L' ¹ kanamycin for transposon insertion strains) in 96-well flat-bottom polypropylene plates (Greiner Bio-One ref: 655261) and incubated them at 30 °C overnight shaking at 800 rpm.

[0168] The following morning, we back-diluted 30 pL from each well into culture tubes containing 3 mL of LB (with 50 mg L' ¹ kanamycin for transposon mutants). Cultures were incubated at 30 °C until they reached an optical density (OD590) of between 1.3 and 1.45. [0169] Each culture was used for 4 biosorption experiments in each of the different conditions detailed in Table 1. Each culture was split into two 1.7 mL centrifuge tubes, pelleted at 7,800 x g, resuspended in 1 mL of buffer, and then pelleted one more time at 7,800 x g. The first tube was rinsed with a low ionic strength buffer (20 mM MES, 20 mM NaCl, adjusted to pH 5.5 with 5M NaOH) and the second tube with a high ionic strength buffer (20 mM MES, 100 mM NaCl, adjusted to pH 5.5 with 5M NaOH).

[0170] We resuspended the rinsed cells in 600 pL of the same respective buffer, took the optical density (OD), then divided each culture into two new tubes with a final OD of 0.85, a volume of 400 pL, and either high or low REE concentrations as follows. For cultures in low ionic strength solution (final concentrations of 10 mM NaCl and 10 mM MES), the low REE solution contained 30 pM each of lanthanum, europium, and ytterbium and the high REE solution contained 60 pM of each. For cultures in high ionic strength solution (final concentrations of 50 mM NaCl and 10 mM MES), the low REE solution contained 10 pM each of lanthanum, europium, and ytterbium and the high REE solution contained 30 pM of each.

[0171] The cultures were incubated for 10 minutes with the REE, prior to a new round of centrifugation (again, 2 minutes at 7,800 x g). We then transferred 300 pL of our supernatant to 0.45 pm Supor 96-well filter plates (Pall Corporation ref: 8029). We previously found that these plates adsorb at most a few percent of the REE in solution. Additionally, we expect the filters to be saturated with REE and thus our estimate for the total amount of biosorbed REE does not depend on how much REE was adsorbed by the filter. From the filtered supernatant, we prepared our ICP-MS samples as described below.

[0172] ICP-MS Measurements

[0173] The ICP-MS samples were prepared by diluting our biosorption samples 1/25 in 2% trace metal grade nitric acid (JT9368, J.T. Baker, Radnor, PA). Our samples were analyzed using an Agilent 7800 ICP-MS (m/z: La, 139; Eu, 151; Yb, 172) using a rare earth element mix standard which included all the other rare earth elements in addition to the three we analyzed (67349, Sigma-Aldrich, St. Louis, MO) and a rhodium in-line internal standard (SKU04736, Sigma-Aldrich, St. Louis, MO, m/z = 103). ICP-MS data were analyzed using the program MassHunter, version 4.5. Quality control was conducted by doing periodic measurements (every ten samples) of our standards (the 10, 25, 50, and 100 ppb) and 2% nitric acid blanks. We used the Rh internal standard to account for effects of drift. Repeat standards were analyzed periodically (or every 10 samples) and were quantified with an accuracy of +/- 2.5% [0174] Comparing Transposon Containing and Wild-Type S. oneidensis Strains

[0175] We found that the average biosorption of transposon insertion strains did not resemble our wild-type bacteria. We theorize that even if the gene a transposon was inserted into did not alter biosorption, it is possible that the transposon itself — or the fact that the insertion mutant strains were grown up with Kanamycin — impacts biosorption. To test this hypothesis, we took four transposon mutants (which we refer to as quasi-wild-type or qWT) that had the transposon in a presumably neutral location and whose biosorption in the As-III screen did not significantly differ from average mutant in the containing plate.

[0176] Choice of Quasi -WT Strains

[0177] We expected that a transposon appearing at the end of a gene would have no effect on that gene. We thus picked transposon mutants where the transposon appeared at the very end of the gene. We also ensured that the transposon was at least 300 base pairs away from the start of any other gene to minimize disruption of promoter regions. Finally, we confirmed that the selected disruptions did not have any significant changes in biosorption within our assay. Four transposon mutants were selected at random from the mutants that met these requirements. The end of the genes where the transposon appeared were SO 4279 for qWTi, SO _4707 for qWT ₂ , SO 0214 for qWT ₃ , and SO 2225 for qWT ₄ .

[0178] Biosorption of each qWT was compared to that of the natural WT and the other qWT strains using a two-sided t-test. We found that the wild-type showed at least 13% higher biosorption compared to the average qWT in every solution condition (Figure 8). Additionally, the qWT had solution condition-dependent differences in total biosorption compared to each other. qWT ₄ had higher (p < .05) biosorption than the average qWT for the HH solution condition. When we performed pairwise t-tests between our four qWT mutants, we found that in three out of four of our solution conditions, we had some mutant or mutants that had different biosorption than the others. In LL, qWT? had significantly higher biosorption than each of the other qWTs. In LH, qWT ₂ had higher biosorption than qWT ₃ and qWT ₄ . In HH, qWT ₄ had significantly higher biosorption than every other qWT and qWT ₃ had significantly lower biosorption than qWTi.

[0179] Methodology for Comparing Relative REE Biosorption

[0180] We compared the amount of biosorption for each individual REE to the total REE biosorption for each solution condition. We found that, over a finite range of total REE biosorption, there was generally a linear relationship between individual and total REE biosorption. We thus used our data for our transposon insertion mutants to plot lines of best fit in each solution condition comparing each individual REE biosorption to total REE biosorption. We expected the baseline calculated from these mutants to reasonably resemble the true average of S. oneidensis transposon insertion strains.

[0181] We excluded our two biggest disruption mutant total biosorption outliers — nusA and dSO 4685 — from our analysis. We left these strains out of our analysis because their total REE biosorption fell outside of the finite range of total biosorption that we felt confident was linear with individual REE biosorption.

[0182] Once we had our line of best fit, f REE _T f for each data point, we calculated the percent change of biosorption of the individual REE of interest (REEi) on the y-axis compared to the expected value based on the total REE biosorption (REE-f) on the x-axis: f REE _T ) REEi _eac h strain we calculated the mean and standard error of this percent /■(REET) ' change. Significance was calculated by doing a two-sided t-test looking to see if our percent change of biosorption was significantly different than 0.

[0183] Effects of Extra Incubation Time on Biosorption

[0184] Previous research has found that, in a sufficiently high pH environment, the longer bacteria were exposed to REE, the less biosorption occurred ⁴² . The authors theorized that this decrease in biosorption over time was caused by bacterial secretions. Other changes, such as changes to bacterial viability within a non-optimal media, could also be responsible for these time-dependent changes. Since they found that this occurred at pH 5.8 and we conducted our assays at pH 5.5, we conducted experiments to test how much different incubation times could affect r results. Since the amount of time we let our bacteria mix with REE was fixed inside our experiment, we chose instead to conduct our experiments by changing the amount of time we let our bacteria sit after rinsing was completed, but prior to adding the REE for our assays.

[0185] We tested the effects of secretions in two of our biosorption environments, LL and HH. We did not find a statistically significant impact on absolute biosorption or on the separation factor for LL. While our results were not statistically significant, it did appear like there was a clear downward trajectory to the level of biosorption as well as an increase in the Yb/La and Eu/La separation factors. For HH, on the other hand, there was a substantial decrease in the overall biosorption level as well as a substantial increase in the Yb/La and Eu/La separation factors. However, the overall biosorption decrease between the 74 minute and 138 minute waiting periods lacked statistical significance and was far less than the decrease between the 42 and 74 minute measurements. This suggests that the effects of the secretions (or whatever other mechanism is responsible for the change in REE binding) decrease over time — a result that would provide confidence given that the bacteria typically sat in their final assay solutions for between 90 and 120 minutes prior to the completion of the biosorption assay.

[0186] Statistical Information [0187] Statistics relating to the genetic screen (identifying outliers, gene ontology enrichment analysis, operon enrichment analysis) are all described in their respective sections. All other statistics were performed using two-tailed t-tests with three biological replicates each.

[0188] Table SI. 29 Genes that control Eu-biosorption are also regulated by the Arc system. We found 29 genes whose disruption changes Eu-biosorption amongst the 604 genes controlled by the Arc system [Lassak2013a], If gene expression is increased by deletion of arcS (measured in Lassak et al. [Lassak2013a]), this suggests that the gene is repressed (H) by the Arc system. Alternatively, if expression goes down, the gene is activated (— ) by Arc. If gene knockout reduces biosorption relative to wild-type, then the gene promotes REE biosorption (]'). Likewise, if the knockout increases biosorption, then the gene discourages it (J,). We speculate that all of these effects are relative to a moderately hypoxic environment since liquid in microwell plates used in high-throughput screening is typically not well mixed. We envision four scenarios in which disruption of the Arc system (by disruption of hplA) can affect biosorption: (1) de-repression of genes that promote REE biosorption; (2) a failure of activation of genes that discourage REE biosorption; (3) derepression of genes that discourage REE biosorption and (4) failure of activation of genes that promote biosorption. The increase in biosorption by disruption of hptA suggests to us that the combined effects of scenarios 1 and 2 dominate the combined effects of scenarios 3 and 4. [0189] Table S2. ICP-MS measurements validate the results of high-throughput Eu-biosorption screening in up to 79% of cases. This table is a complement to Figures 3A-H in the main text. Twenty-five genes highlighted by high-throughput screening with the Arsenazo-III (As-III) assay were selected for further analysis by mass spectrometry in four solution conditions (detailed in Table 1 in the main text): low ionic strength, low total initial REE (LL); low ionic strength, high total initial REE (LH); high ionic strength, low total initial REE (HL); and high ionic strength, high total initial REE (HH). H: higher biosorption than quasi -wild-type; L: lower total lanthanide biosorption than quasi-wild-type; NSC: no significant change. PSI : Polysaccharide and O-antigen Synthesis Operon 1; PS2: Polysaccharide and O-antigen Synthesis Operon 2.

[0190] Table S3. Summary of projected changes to length of REE separation process caused by dwbpA mutant. This table summarizes the results of Figure 12. The baseline is what our average mutant bacterium REE biosorption distribution would look like if it biosorbed the same total amount of REE as dwbpA (see Figure 4E for details about this baseline).

[0191] Supplementary Notes [0192] Note SI. Theory of REE-separation by Biosorption and Desorption

[0193] Selective biosorption enables separation of metals by splitting a mixture of metals into a bound fraction (for instance enriched in one or more of the metals) and a free fraction (depleted in one more of the metals). The bound and free fractions can be physically removed from one another, enabling separation of a target metal. While the effect of the selective biosorption on the purity of a target metal (say Eu) is small in any individual step, the effect of successive enrichment is not.

[0194] We use a model shown in Figures 10A-D to illustrate the effects of biosorption selectivity on the separation of three lanthanides (Mi, M2 and M3; e.g., Eu, Yb and La). For our example, we have chosen Eu as the target metal to purify as wild-type S. oneidensis has a slight preference for it under low ionic strength conditions (see Figures 3A- H in the main text). The separation system consists of a chromatographic columns containing immobilized biomass (e.g, on a filter [Bonificio2016a]; a biofilm on a solid support; or encapsulated in gel beads [Brewer2019b]).

[0195] The system is loaded with a solution initially containing a mixture of metals (e.g, Eu, Yb and La). After equilibration the free fraction (the liquid) is removed from the column and moved to a wash collection container. Next, the bound fraction is eluted (for example by a pH swing [Bonificio2016a, ParkD2020a]). The eluant is then pH adjusted (to compensate for the pH swing), and its volume adjusted (to compensate for any differences in the loading and elution volume), where it can be loaded into the same or a different column. This process of load, bind, elute and re-load is repeated until the desired purity of Mi in the eluant is achieved ( wi,b = 0.9, 0.95, or 0.99). At each step, the amount of biomass is adjusted so that half of the metals loaded into the column are bound, while the other half are left free in solution.

[0196] To model the separation process, we use a system of 3 simultaneous equations with 3 unknowns. The unknown concentration of metals that are free in solution in the column (CEU, f, CYb, f, and CLa, f) and the concentrations of metals bound to the biomass (CEU, b, CYb,b, and CLa, b); the known separation factors for each of the pairs of metals (af^ and a^); and the known analytical concentrations of the metals ( EU, T, CYb, T, and La, T).

[0197] The separation factors (measured) can be related to the free and bound concentrations of the metals (both unknown), aEu ⁼ . ^c Yb,b/ ^c Yb,f ^> )/ . ^c Eu,b/ ^c Eu,f ^> )’ (S2) The free concentrations of the metals can be expressed as a function of the known analytical concentrations of the metals and the bound concentrations,

Thus, Equations SI to S3 can be re-cast and solved in terms of just the three unknown bound concentrations of metals,

Eu ^c Eu,b(, ^c La,T~ ^c La,b) a La (54) cLa,b ( ^c Eu,T~ ^c Eu,b )

Yb > ^c Yb,b( ^c Eu,T~ ^c Eu,b) (55)

CEu,b cYb,T~ ^c Yb,b)

Furthermore, we add as a constraint that there are sufficient binding sites to bind half of the total metals,

In the case that S. oneidensis acts as if it has a single type of binding site, the separation factors can be reduced to the ratios of the dissociation constants of the site for the three metals. For example, where CB,L is the concentration of free binding sites. In this case, the separation factor is independent of the composition of the loaded solution, and hence remains constant throughout the separation process.

[0198] After equilibration, the solution phase is removed from the column and the bound phase is eluted. The bound phase is then reloaded into the column. The number of binding sites is reduced so that it is equal to half the total number of metals loaded.

[0199] Equations S4 to S6 are numerically solved with the separation factors for the dwbpA mutant using a code in the REE-SELECTIVITY package [Medin2023a], the purity of Eu is calculated at each biosorption/elution cycle (Figure 12), and the number of steps to reach target purities of 95, 99, and 99.9% purity are tabulated in Table S3.

[0200] Table 1. Solution environments for detailed REE biosorption measurements.

All solutions were adjusted to pH 5.5.