Selective bacterial separation of critical metals: towards a sustainable method for recycling lithium ion batteries

The large scale recycling of lithium ion batteries (LIBs) is essential to satisfy global demands for the raw materials required to implement this technology as part of a clean energy strategy. However, despite what is rapidly becoming a critical need, an efficient and sustainable recycling process for LIBs has yet to be developed. Biological reactions occur with great selectivity under mild conditions, offering new avenues for the implementation of more environmentally sustainable processes. Here, we demonstrate a sequential process employing two bacterial species to recover Mn, Co and Ni, from vehicular LIBs through the biosynthesis of metallic nanoparticles, whilst Li remains within the leachate. Moreover the feasibility of Mn recovery from polymetallic solutions was demonstrated at semi-pilot scale in a 30 L bioreactor. Additionally, to provide insight into the biological process occurring, we investigated selectivity between Co and Ni using proteomics to identify the biological response and confirm the potential of a bio-based method to separate these two essential metals. Our approach determines the principles and first steps of a practical bio-separation and recovery system, underlining the relevance of harnessing biological specificity for recycling and up-cycling critical materials.


Methodology
Proteomics analysis was designed based on the removal rates of Co and Ni and considered the following factors: 1) three different metal combinations, single metal treatments (only Co 2+ or Ni 2+ ) and bimetallic (Co 2+ and Ni 2+ at equal concentrations) to mimic the removal experiments; 2) two different metal concentration levels (10 and 100 ppm) and 3) two different time points (2 and 20 h) to investigate the dynamics of proteins production. D. alaskensis G20 cultures for proteomic analysis were grown and washed in MOPS buffer similarly to the metal removal studies and previous proteomics analysis with this organism. 1 After 2 and 20 h incubation cell cultures were collected by centrifugation, the supernatants discarded and cell pellets stored at -75 °C until proteomic analysis. ESI-HPLC-MS/MS analysis Samples were reconstituted in 8 M urea and protein concentration was determined by Bradford protein assay (Biorad). 30µg of samples were digested using S-Trap TM (ProtifFi) following manufacturers protocol. After speedvacuum drying, peptide samples were re-suspended in MS-loading buffer (0.05% v/v trifluoroacetic acid in water) to 1µg µl -1 final concentration and then filtered using Millex filter before HPLC-MS analysis. 5µl of a 1 to 1 dilution (in 0.05% TFA) was injected for analysis. Nano-ESI-HPLC-MS/MS analysis was performed using an online system, the nano-HPLC (Dionex Ultimate 3000 RSLC, Thermo-Fisher) coupled to a QExactive mass spectrometer (Thermo-Fisher) with a 300µm x 5 mm pre-column (Acclaim Pepmap, 5µm particle size) joined with a 75 µm x 50 cm column (EASY-Spray, 3 µm particle size). The nano-pump was run using solvent A (2% Acetonitrile in water 0.1% formic acid) and solvent B (80% acetonitrile, 20% water and 0.1% formic acid) and peptides were separated using a multi-step gradient of 2-98% buffer B at a flow rate of 300 nl min -1 over 90 minutes.
Data process and analysis Progenesis (version 4 Nonlinear Dynamics, UK) was used for LC-MS label-free quantitation and data normalisation and analysis. Filtering was carried out so that only MS/MS peaks with positive charges of 2, 3 or 4 were taken into account for the total number of 'features' (signal at one particular retention time and m/z) and only the five most intense spectra per 'feature' were included. MS/MS spectra was searched using MASCOT Version 2.4 (Matrix Science Ltd, UK) against a custom D. alaskensis G20 database with maximum missed-cut value set to 2 as in previous proteomics work with this organism. 1 For convention the protein identifier (protein ID) used in the present work has been given the name of its encoding gene which starts with letters "Dde_" followed by four numeric digits. The updates related to genes and protein annotations can be found at the KEGG database (https://www.genome.jp/kegg/). From the Progenesis exported results sheet, differentially expressed proteins were considered significant if the p-value was less than 0.05 (ANOVA) and if the number of peptides used in quantitation per protein was equal to or more than 2. Heat maps were created using R (https://cran.r-project.org/) after log10 transformation of the normalised abundance data sets.

Proteomics results
To fully develop our understanding of a biological separation process we investigated the biological molecules and mechanisms responsible for Co and Ni nanoparticle synthesis. The proteins produced by D. alaskensis G20 after incubation with Co 2+ and/or Ni 2+ were analysed by Electrospray Ionisation Mass Spectrometry (ESI)-HPLC-MS/MS. 1579 proteins were identified in this study which corresponds to 52% of the total proteome of D. alaskensis G20, 2 providing comprehensive coverage of the bacterial response to the presence of these two transition metals, relative to previous studies. 3,4 After 2 h incubation there was a remarkable reduction in the protein abundance in treatments containing 100 ppm of Ni 2+ in comparison to the control. It is also particularly noteworthy that metal-binding proteins, such as the UPF0173 metal-dependent hydrolase (Dde_0151), the quinone-interacting membrane-bound oxidoreductase (Dde_1113) and the MJ0042 family finger-like protein (Dde_1116), showed the highest abundance in the presence of 100 ppm Co 2+ and yet the lowest abundance for the 100 ppm bimetallic treatment (Fig. S3B). UPF0173 metal-dependent hydrolases and MJ0042 family finger-like proteins are known for binding Zn 2+ that can be exchanged by Co 2+ without a loss of functionality. 5,6 To reduce the risk of mis-metallation with Ni, or perhaps to remedy such, it is understandable that a decrease in their abundance is observed and this might then also explain why Co removal from the dissolved fraction dropped when Ni 2+ was present at concentrations ≥ 50 ppm. After 20 h incubation, 23 proteins showed higher abundance in the treatment with Ni 2+ compared to the Control and Co 2+ treatments (Fig. S3C). Some of these proteins are likely responsible for mediating Ni 2+ toxicity, such as the zinc resistanceassociated protein (Dde_0111) which belongs to a family of four-helix hooked hairpins. 7 Also proteins that might be involved in metal reduction processes such as oxidoreductases, 8,9 were significantly more abundant after 20 h incubation with 10 ppm of Ni 2+ compared to the treatment with Co 2+ , confirming a distinctive cellular response depending on the metal. Some of these oxidoreductases, such as the FAD/NAD (P)-binding domain protein (Dde_1381) and the FAD-dependent pyridine nucleotidedisulfide oxidoreductase (Dde_2176), classified within the xenobiotics biodegradation and metabolism pathways, might be responsible for reducing Ni 2+ into a less toxic form of this metal.
We identified similar cellular responses to Co 2+ , and Ni 2+ , with regard to ABC transporters as the periplasmic component of zinc ABC transporter protein (Dde_2208) and the Molybdenum ABC transporter (Dde_0155), were both more abundant in the presence of any metal treatment compared to the control. Dde_0155 had been identified in previous work with D. alaskensis G20 after incubation with Pd 2+ and Pt 4+ 1 suggesting that this protein is a key component of heavy metal detoxification pathways. 10 Conversely, other proteins related to the ABC transporter pathways were found at significantly larger concentrations after incubation with Co 2+ alone than after treatments containing Ni 2+ . These were two periplasmic subunit family 3 proteins, Dde_0168 (Fig. S3A-B) and Dde_1429 (Fig. S3A) related to export mechanisms and the cell division ATP-binding FtsE protein (Dde_0114) (Fig. S3A-B). 1

Fig. S3 Proteomics analysis. Heat maps of proteins abundance associated to the control and metal treatments (Co, Ni and bimetallic) after 2 h incubation with metals at (A) 10 ppm and (B) 100 ppm and (C) after 20 h incubation with metals at 10 ppm.
Values are represented as the log 10 (Mean protein abundance, n=3 biological replicates). Rows are labelled with the protein identifier Dde_#### followed by the *metal treatment exhibiting the highest significant abundance (ANOVA, p-value <0.05, number of peptides detected ≥ 2, absolute ratio abundance > 1.5-fold). The protein abundance, ANOVA results and full name of the proteins together with their associated ID is available at https://doi.org/10.7488/ds/3130.