A phylogenetic tree was generated using the maximum likelihood method in MEGA X version 10.0.5, based on multiple sequence alignment of M38 β-aspartyl peptidase from Fervidobacterium islandicum AW-1 (FiBAP) and homologs sharing ≥ 40% amino acid sequence identity. A bootstrap consensus tree was inferred from 1,000 replicates. Evolutionary distances were computed using the Poisson correction method and expressed as the number of amino acid substitutions per site. Amino acid sequence alignment was performed using ClustalX software V2.1. Sequence similarities and secondary structure information from aligned sequences were generated using ESPript 3.0 (Robert and Gouet, 2014).

Genomic DNA was isolated from F. islandicum AW-1, purified using a genomic DNA extraction kit according to the manufacturer's instructions (Qiagen, Hilden, Germany), and then used as template for PCR amplification. The NA23_RS08100 gene encoding FiBAP was amplified by PCR using forward (5′-GCTAGCATGATAAAAATTATAAAGAACG-3′) and reverse (5′-CTCGAGTCAAAATTCAAAGTTTAAGTTC-3′) primers (underlined sequences represent the restriction sites for NheI and XhoI). The PCR product was digested with NheI and XhoI and cloned into the pET-28a(+) vector (Novagen, San Diego, CA), yielding pET-FiBAP. The resulting plasmid encodes the target gene fused to a 6 × His-tag at the N-terminus. For expression of recombinant FiBAP, E. coli BL21 (DE3) cells transformed with pET-FiBAP were grown at 37°C in 1 L of Luria-Bertani (LB) medium containing kanamycin (50 μg/mL) to an optical density at 600 nm (OD₆₀₀) of 0.5–0.8. After induction with 1 mM Isopropyl-1-thio-β-D-galactopyranoside (IPTG), cells were grown overnight at 37°C, then harvested by centrifugation (10,000 × g, 20 min, 4°C) and stored at −80°C.

The harvested cells were resuspended in 20 mM Tris-HCl buffer containing 500 mM NaCl, 5 mM imidazole, and 1 mM phenylmethylsulfonyl fluoride (PMSF) (pH 7.9) and disrupted by sonication. After centrifugation at 10,000 × g for 20 min, the supernatant was heated at 60°C for 30 min then centrifuged at 10,000 × g for 20 min to remove denatured E. coli proteins. The supernatant was applied to Ni²⁺-affinity resin (10 mL) equilibrated with the same buffer, and His-tagged protein was eluted with 20 mM Tris-HCl buffer containing 500 mM NaCl and 250 mM imidazole. Samples were loaded on a Superdex 200 10/300 GL column (GE Healthcare, USA) equilibrated with 25 mM Tris-HCl buffer (pH 7.5) containing 150 mM NaCl (Supplementary Table 1). The major fractions were analyzed by 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and visualized using Coomassie Blue staining.

To monitor the recombinant FiBAP, western blotting was performed using a 6 × His tag monoclonal antibody (ThermoFisher Scientific, USA). Each sample was subjected to 12% SDS-PAGE, electrophoretically transferred to a polyvinylidene fluoride (PVDF) membrane (Bio-Rad, Hercules, CA, USA), blocked with 5% skim milk in TBST buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.05% Tween 20) at 4°C overnight, and incubated with horseradish peroxidase-conjugated 6 × His epitope monoclonal antibody (1:1,000 dilution) for 2 h at room temperature. The membranes were then washed three times with TBST buffer for 10 min and were developed using a WESTSAVE Up western blotting detection system (AbFrontier, Seoul, Korea).

Recombinant feather keratins were prepared as described previously (Jin et al., 2017) and used as native substrates. Briefly, E. coli cells expressing recombinant feather keratins were resuspended in lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, 1 mM PMSF, pH 8.0) and disrupted by sonication. After centrifugation at 10,000 × g for 30 min, expressed keratins in the form of inclusion bodies were resuspended in lysis buffer containing 8 M urea and 1 mM PMSF, incubated on ice for 1 h, and centrifuged at 16,000 × g for 30 min. Supernatants were filtered through a 0.45 μm membrane, then applied to a 10 mL Ni-NTA agarose resin column (Qiagen, Germany) equilibrated with lysis buffer containing 8 M urea. Fractions containing unfolded keratins were eluted with 250 mM imidazole, concentrated using an Amicon Ultra-3K device (Millipore, USA), and buffer-exchanged by step-wise dialysis against 50 mM Tris-HCl (pH 8.0) at 4°C. Dialyzed samples were centrifuged at 10,000 × g for 30 min to remove insoluble material, and the resulting supernatants containing refolded keratins were concentrated using an Amicon Ultra-3K device (Millipore).

FiBAP activity was determined by measuring the increase in free amino acids (Rosen, 1957). Reaction mixtures (0.16 mL) containing 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer (pH 8.0), 500 ng/mL enzyme, and 1 mM β-Asp-Leu (β-DL) as a synthetic substrate were incubated at 80°C for 15 min. The reaction was stopped by adding trichloroacetic acid (TCA) reagent. After an additional 15 min incubation at 100°C for color development, followed by cooling on ice, the absorbance was measured at 570 nm. One unit of FiBAP activity was defined as the amount of enzyme that produced 1 μmol of product per min under the assay conditions (U/mg). Additionally, to assess the proteolytic activity of the purified enzyme, we measured the increase in free amino acids using the ninhydrin assay (Rosen, 1957) with casein and gelatin as substrates.

The temperature dependence of FiBAP activity was measured using the standard protocol but the reaction temperatures was varied from 40 to 98°C. To determine the effect of pH on FiBAP activity, reaction mixtures (0.16 mL) were incubated at 80°C under standard assay conditions but HEPES buffer was replaced by 50 mM sodium acetate buffer (pH 4.0–6.0), HEPES buffer (pH 6.0–8.0), borate buffer (pH 8.0–10.0), or sodium bicarbonate buffer (pH 9.0–10.0). All pH values were adjusted at room temperature, and ΔpK_a/ΔTs (in which the latter term is the change in temperature) for each buffer was taken into account when the results were analyzed.

To determine the kinetic parameters, assays were performed in 50 mM HEPES (pH 8.0) containing 1 mM CoCl₂, 500 ng/mL enzyme, and 1–32 mM β-DL substrate. Reaction mixtures were incubated for 10 min at 80°C and stopped by cooling on ice. Kinetic parameters were obtained by fitting the empirical data to the Michaelis-Menten equation using Origin 8.0 software.

To analyze the effect of metals, purified FiBAP was treated with 20 mM EDTA at room temperature for 2 h to remove metal ions, followed by overnight dialysis against 20 mM Tris-HCl (pH 7.5) at 4°C with three changes of buffer. The divalent metal ion content of both as-isolated and EDTA-treated samples were determined by high-resolution inductively-coupled plasma mass spectrometry (ICP-MS) on a PlasmaQuad 3 instrument. To assess the effects of various metal ions on FiBAP activity, the EDTA-treated samples were preincubated with 1 mM CoCl₂·6H₂O, MnCl₂·4H₂O, MgCl₂·6H₂O, CaCl₂·2H₂O, ZnCl₂·6H₂O, CuCl₂·2H₂O, FeCl₂·6H₂O, or NiCl₂·6H₂O for 15 min at 50°C. The residual activity was measured in triplicate under standard assay conditions.

Preliminary crystallization trials were performed using purified FiBAP (10 mg/mL) in an appropriate buffer (20 mM Tris-HCl pH 7.4, 50 mM NaCl) mixed with protein crystallization solutions (Hampton Research and Wizard; 1:1 volume ratio) by the hanging drop vapor-diffusion method at 20°C (293 K) as described previously (Lee et al., 2015b). The initial FiBAP crystals were further improved by modifying the Crystal Screen Lite 34 conditions (0.1 M sodium acetate pH 4.6 and 1 M sodium formate). FiBAP was co-crystallized with substrate analog N-carbobenzoxy-β-Asp-Leu (Cbz-β-DL) at a molar ratio of 1:1.2, and crystals of the complex were obtained in modified PEG/Ion screen 2–32 conditions (2% v/v tacsimate pH 5.0, 0.1 M sodium citrate tribasic dihydrate, pH 5.6; 16% w/v polyethylene glycol 3350). Crystals were soaked for 30 s in a cryosolution containing mother liquor plus 20% glycerol, then frozen in liquid nitrogen prior to synchrotron radiation diffraction experiments.

X-ray diffraction data were collected from FiBAP crystals on beamline 7A at the Pohang Light source (Pohang, Korea) using an ADSC Q270 detector, with an oscillation of 1.0° and a 1 s exposure per frame over a 360° range at a wavelength of 0.97934 Å. Crystals of FiBAP with and without β-DL as substrate grown using the above conditions diffracted to 2.6 Å and 2.7 Å, respectively. Diffraction data were processed using the HKL-2000 program (HKL Research Inc.). In the case of the ligand-free structure, molecular replacement (MR) was performed with EcIadA (PDB: 1ONW) as the search model using the CCP4 program MOLREP (Vagin and Teplyakov, 2010). The initial model was further refined using the PHENIX (Afonine et al., 2010) and REFMAC5 (Murshudov et al., 2011) programs to achieve a model with R_work and R_free values of 16.9 and 22.1%, respectively. In the case of the Cbz-β-DL-bound FiBAP diffraction dataset, to enhance the overall data quality, two native datasets of equivalent diffraction limits were combined using the Scale_and_Merge function of the PHENIX suite (Adams et al., 2010). MR was performed for the Cbz-β-DL-bound FiBAP structure with the refined crystal structure of ligand-free FiBAP as the search model using the Phaser program (McCoy, 2007). The chemical coordination file for β-DL was built using Coot (Emsley and Cowtan, 2004) and eLBOW (Moriarty et al., 2009). The ligand was fitted into the map, and several rounds of refinement were performed using the REFMAC5 and PHENIX programs, yielding R_work and R_free values of 19.7 and 27.5%, respectively. The structures of FiBAP with and without Cbz-β-DL have been deposited in the Protein Data Bank (PDB) under accession codes 7CDH and 7CF6, respectively.

The leucine (Leu) auxotrophic E. coli BL21 (DE3) strain was constructed by deleting the leuB gene encoding 3-isopropyl malate dehydrogenase, which is involved in leucine biosynthesis, by the RED/ET recombination method using a Quick & Easy E. coli Gene deletion kit (Gene Bridges) according to the manufacturer's instructions. A functional cassette flanked with FRT and homology arms was generated by PCR using the FRTPGK-gb2-neo-FRT fragment as a template and primers 5′-AGTTGCAACGCAAAGCTCAACACAACGAAAACAACA-AGGAAACCGTGTGAAATTAACCCTCACTAAAGGGCGG-3′ (forward) and 5′-GTACACAACGTGAGCGTCGAACAAT-TTTTCGTATAACGTCTTAGCCATGATAATACGACTCACT-ATAGGGCTCG-3′(reverse). The kanamycin selection marker was removed by transforming cells with FLP recombinase expression plasmid 706-FLP (Gene Bridges). After selecting mutants lacking the selection marker cassette by streaking each colony on LB agar plates with and without 50 μg/mL kanamycin, deletion of the leuB gene was confirmed by PCR with primers 5′-GCTAACTACAACGGTCGCCGCTTCCACGGCGTC-3′ (forward) and 5′-GGCGCGCAGACCATCGAACGCCTGCGG-TGAGG-3′ (reverse).

For functional complementation of FiBAP, a functional cassette flanked with FRT and homology arms was generated to delete the iadA gene encoding EcIadA using primers 5′-CATTGCTGTCGATCTGGGTTATGCAGCTTATTGTTTAA-CAAGGAGTTACCAATTAACCCTCACTAAAGGGCGG-3′(forward) and 5′-TCAGCCGCCCTTGCGGGCATTCTACG-TCCATTCGGGCGGCTGACAACCGTTAATACGACTCACT-ATAGGGCTCG-3' (reverse). In addition, primers 5′-GTC-GGTCGCTGCCTGGGGACAGCCGAAGTG-3′ (forward) and 5′-CTGAGAGTTGCAGCGGCGTTACCTGGCGG-3′ (reverse) were used to confirm the iadA deletion mutants.

To monitor the growth phenotypes of the Leu auxotrophic E. coli strains harboring the pET28a-FiBAP and the pET28a-iadA plasmids, each protein-overexpressing strain was streaked on appropriate minimal M9 agar plates containing 50 μg/mL kanamycin and 1 mM IPTG supplemented with different compositions of amino acids (19 amino acids with and without Leu) in the presence and absence of 1.6 mM β-DL. To investigate the effect of heat shock on bacterial growth retardation, each strain was grown at 37°C to an OD₆₀₀ of 1.2–1.5 in LB medium containing kanamycin (50 μg/mL), and then incubated at 44°C for 0.5 h or longer until cells entered the stationary phase.

To determine the total protein content of soluble (supernatant) and insoluble (pellet) fractions, cells were harvested before (mid-exponential phase) and after (stationary phase) heat treatment. The same amount of cells, as judged by the OD₆₀₀ value, was resuspended in 50 mM Tris-HCl buffer (pH 8.0) and disrupted by sonication. After centrifugation at 10,000 × g for 30 min, the protein concentration of fractionated samples was determined using the bicinchoninic acid (BCA) assay (Smith et al., 1985), and each subcellular fractions was visualized on the 12% acrylamide gel for SDS-PAGE analysis.

Keratin hydrolysates were analyzed by reversed-phase HPLC-ESI-MS/MS using a NanoLC-2D Ultra system (Eksigent, Dublin, CA, USA) coupled to a LTQ-XL mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) in direct injection mode. Briefly, after a 5 μL injection, keratin hydrolysates were loaded onto a reversed-phase ProteoPepII C18 column (5 μm, 300 Å pore size, 0.15 × 25 mm; New Objective IntegraFrit, Scientific Instrument Services, Inc., Ringoes, NJ, USA) and eluted onto a Molex Polymicro Flexible Fused Silica Capillary Tubing (I.D. 75 μm, O.D. 375 μm) at a flow rate of 0.4 μL/min. The gradient consisted of mobile phase A [0.1% formic acid (v/v) in water], mobile phase B [0.1% formic acid (v/v) in acetonitrile], 2% solvent B (0–1 min), 5% solvent B (1–62 min), 35% solvent B (62–65 min), 60% solvent B (65–70 min), and 2% solvent B in A (70–90 min) with a total runtime of 90 min, including mobile phase equilibration.

MS analysis of peptide eluents was performed on a LTQ-XL system (ThermoFisher Scientific) in positive-ion mode with a nano ion spray voltage of 2.3 kV, a scan range of 300–1,800 (m/z), a curtain gas pressure of 20 psi, a nebulizer gas pressure of 6 psi, and an interface heater temperature of 200°C. Full-scan MS spectra were acquired from 35 precursors selected for MS/MS analysis from the 300 to 1,800 m/z range, utilizing a dynamic exclusion of 30 s. The IDA collision energy (CE) parameter script, selecting up to 35 precursors with charge states of +2 to +3, was employed to automatically control the CE.

Data were processed using MSConvert (ProteoWizard). This software converts raw data (.RAW format) into peak lists (.mgf format). The FASTA database employed contained the Gallus gallus keratin sequence, and this afforded the opportunity to employ the target decoy database search strategy (Elias and Gygi, 2007). Data containing both MS and MS/MS information were uploaded into Xcalibar software (ThermoFisher Scientific) and used to generate MS-extracted ion chromatograms (XICs) for each identified peptide. For the MS-GF search, carbamidomethyl (C) was set as a fixed modification and oxidation (M) was set as a variable modification. The peptide tolerance was ±4.0 Da, and the MS/MS tolerance was ±1 Da. The software algorithm simultaneously searched all modifications listed in UniMod (http://www.unimod.org/) (Shilov et al., 2007). False discovery rate (FDR) analysis was also performed using integrated tools in PeptideShaker (CompOmics), which generated.mgf files that were subsequently searched against the current G. gallus keratins SwissProt database using MS-GF.

Calorimetric measurements were performed using a VP-DSC microcalorimeter (Microcal Inc., GE Malvern, Worcestershire, UK). All scans were run at pH 7.4 in 10 mM potassium phosphate buffer containing 150 mM NaCl, over a temperature range from 10 to 130°C at a rate of 90°C/h. The cell volume was 0.8 mL. Potassium phosphate buffer was used for baseline scans, and apparent melting temperature (T_m) values of FiBAP (25 μM) were determined.

An ISOQUANT Isoaspartate Detection Kit (Promega) was used for the detection of isoaspartate (isoAsp) according to the manufacturer's instructions. Each cell lysate (2 μg) from heat-shocked strains was incubated at 30°C for 30 min in the presence of L-isoaspartyl (isoAsp) methyltransferase (PIMT) and S-adenosyl methionine (SAM) in the reaction buffer, and the stop solution was then added to halt the reaction. The reaction mixture (40 μl) was loaded onto a reversed-phase YMC-Triart C18 column (250 × 4.6 mm I.D. S-5 μm, 12 nm) equilibrated with 90% 50 mM potassium phosphate (pH 6.2) and 10% methanol. The amount of isoAsp was determined by quantifying eluted S-adenosyl homocysteine (SAH).

To predict the functional role of the NA23_RS8100 gene product, we performed a BLASTP search using the deduced amino acid sequence to identify homologs. Multiple sequence alignment of the NA23_RS8100 gene product with β-aspartyl peptidases (BAPs) revealed that its amino acid sequence shares high levels of sequence identity with BAP from Fervidobacterium species (≥65%) and other genera including Thermosipho (≥55%), Pseudothermotoga (≥51%), Thermotoga (≥51%), Clostridium (≥49%), Bacillus (≥47%), and Enterococcus (≥49%; Figure 1 and Supplementary Figure 1). In addition, the deduced amino acid sequence of the NA23_RS08100 gene product shares pronounced sequence identity with mesophilic BAPs from Escherichia coli (EcIadA; 39% sequence identity; PDB:1ONW) and Colwellia pshchrerythraea 34H (CpIadA; 41% sequence identity; PDB:5XGW), for which crystal structures are available. Remarkably, all BAP homologs derived from extremophiles belong to the Type I isoaspartyl dipeptidase (IadA) family, members of which possess a conserved Glu residue for metal binding in the active site (Supplementary Figure 1). Moreover, phylogenetic analysis showed that BAP from F. islandicum AW-1 (FiBAP) has a highly close evolutionary relationship with homologs from extremophilic Fervidobacterium and Thermosipho, whereas extremophilic FiBAP has diverged away from the Type II IadAs of mesophiles in the order Enterobacteriales and further away from EcIadA and CpIadA members of the BAP family (Figure 1). These results led us to tentatively conclude that the protein encoded by the NA23_RS8100 gene in F. islandicum AW-1 is β-aspartyl peptidase (subsequently named FiBAP), representing an ancient form of IadA that has diverged from Type II IadA.

We successfully expressed the NA23_RS08100 gene as a C-terminal hexahistidine (6 × His)-tagged fusion protein in E. coli BL21 (DE3) cells. To facilitate subsequent purification, E. coli lysates were heat-treated to remove most of the endogenous proteins, and the resulting supernatants containing soluble FiBAP were applied to a Ni²⁺-affinity column. The purified 6 × His-tagged FiBAP fusion protein yields an intact recombinant protein with an estimated M_r of 42,000 by SDS-PAGE (Figure 2A). However, size exclusion chromatography using a Superdex 200 column suggested that the native form of FiBAP is as an octameric protein with an apparent M_r of 318,000 (Figure 2B).

Unlike typical endo-proteases, FiBAP displayed minimal activity toward casein and gelatin as substrates, but it was active against β-Asp-Leu (β-DL), with an apparent optimal temperature of 80°C (Figure 2C) and an apparent optimum pH of 7.0 at 80°C (Figure 2D). To further assess the structural stability of FiBAP, we used DSC to determine the T_m, revealing a major midpoint of the thermal transition (T_m = 96.8 ± 0.5°C) and a minor one at 100.4 ± 1.44°C (Figure 2E).

Since EcIadA is a metalloprotease, we performed ICP-MS analysis of as-isolated FiBAP, demonstrating that it contained Zn²⁺ (1.14 ± 0.37 mol of metal per mol of monomer). To investigate the effects of inhibitors on FiBAP activity, as-isolated enzyme was pre-incubated with various inhibitors and its residual activity was measured. Dithiothreitol, PMSF, and Pepstatin A had little inhibitory effect on FiBAP, but 1% SDS and 1 mM EDTA inhibited enzyme activity (Figure 2F). Subsequently, after a 15 min preincubation of EDTA-treated and dialyzed FiBAP with various metal ions at 50°C, the residual activity was measured under standard assay conditions, demonstrating that FiBAP activity was significantly increased in the presence of Zn²⁺ and Co²⁺ compared with that in the absence of divalent metal ions, suggesting that this enzyme may be a metalloenzyme.

Additionally, we performed steady-state kinetics analysis of FiBAP with β-DL as substrate, yielding K_m and V_max values of 5.23 ± 1.59 mg/mL and 286.33 ± 65.7 μM/min, respectively (Figure 2G).

Notably, FiBAP is the most abundant and highly expressed enzyme when F. islandicum AW-1 is grown on feathers under starvation conditions (Kang et al., 2020). Although a MEROPS (http://merops.sanger.ac.uk) database search provided some information on hierarchical, structure-based classification of the peptidases, it did not yield the predicted cleavage pattern for M38 BAP. To further investigate the hydrolysates resulting from FiBAP degradation, we performed LC-MS/MS analysis using soluble keratins as native substrates (Supplementary Table 2). Remarkably, all peptides hydrolyzed by FiBAP matched the keratin sequences derived from G. gallus sequences, indicating that FiBAP can cleave not only the peptide bond between Asp and Leu, but also other α-peptide bonds between hydrophobic, aromatic, and hydrophilic amino acid residues at the C-terminus (Figure 3), as were the cases of the EcIadA and BAP from rat tissue (Dorer et al., 1968; Gary and Clarke, 1995). This result suggests that FiBAP is a beta-aspartyl peptidase with a relatively broad substrate specificity toward α-peptide bonds.

Crystal structures of the ligand-free form of FiBAP and the ligand-bound form bound with N-carbobenzoxy-β-Asp-Leu (Cbz-β-DL) as a substrate analog were determined at resolutions of 2.6 and 2.7 Å, respectively (Table 1). Cbz-β-DL-free and -bound structures belonged to I422 and P22₁2 space groups, with one and four subunits in the asymmetric unit, respectively. The overall structure of each monomer can be divided into two distinct domains; an N-terminal β-sandwich domain (M1–G56; G344–E386) and a C-terminal catalytic domain (L57–K343), that are mainly involved in dimerization and catalysis, respectively (Figure 4A). The β-sandwich domain located at the N-terminus of the molecule mainly comprises nine β-sheets (β1–β6, β17–β19) arranged in two layers with β1, β3, β4, and β5 in one layer and β2, β6, and β17–β19 in the other layer (Figure 4A). Inter-layer interactions may stabilize the β-sandwich domain consisting of loops and β-strands through hydrophobic interactions and a salt bridge between K376 in β19 and E44 in β5. The catalytic domain at the C-terminus of the molecule folds into a (β/α)₈ triosephosphate isomerase (TIM)-barrel motif. Specifically, eight β-sheets (β7–β13) surrounded by nine α-helices (α2–α10) form the central core containing the substrate-binding site harboring two Zn²⁺ ions (Figure 4A).

A structural similarity search using the DALI server revealed that FiBAP is very closely related to EcIadA, a Type II IadA enzyme that has evolved in the direction of allowing post-translational modification (PTM) of the active site for metal binding and catalytic activity (Park et al., 2017). Structural neighbors identified in the FiBAP query structure include IsoAsp dipeptidase, dihydropyrimidinase-related protein, enamidase, allantoinase, and dihydropyrimidinase, all of which are hydrolases. The structural architecture of FiBAP was compared with previously determined structures of homologs (EcIadA and CpIadA). All monomers were superimposed with an average root mean square deviation (RMSD) of 0.705 Å for all Cα atoms (Figure 4B), suggesting that the overall scaffold of FiBAP is similar to that of its homologs. Among these BAPs, structural elements are highly conserved, except for a minor difference in the flexible loop connecting β4 and β5 (Figures 4A,B), which might be of no discernable biological significance.

Similar to EcIadA (Elias and Gygi, 2007) and CpIadA (Shilov et al., 2007), size-exclusion chromatography (Figure 2B) indicated that FiBAP is an octamer in solution (S1–S8 in Figure 4C), usually referred to as a “tetramer of dimers” (Figures 4C,D), as supported by symmetry-related molecule analysis. The initial dimer is formed when the β-sandwich domain of each monomer (Figure 4A) interacts with an adjacent monomer, mainly through hydrophobic interactions involving several residues (I2, I4, V34, V35, V38, L39, F41, and I46) in the loop between β4 and β5 (upper red box in Figure 4D). Additionally, the substrate-binding cavity of the catalytic domain (Figure 4A) participates in dimerization by providing the second point of contact between adjacent partners through salt bridges. These four salt bridges are formed by E107 (α2), K114 (α2), R143 (α3), and D149 (α3) of each monomer of the dimer (lower red box in Figure 4D). Therefore, the hydrophobic interactions between β-sandwich domains, followed by hydrogen (H)-bonds and salt bridges between catalytic domains of adjacent subunits results in dimer formation (Figure 4B and Supplementary Table 3). Remarkably, the corresponding region in structural homologs such as EcIadA and CpIadA does not appear to contain some of these salt bridges because the distances between charged amino acids are >4 Å in all cases except E114-K150 of EcIadA (Supplementary Figure 3B) (Thoden et al., 2003; Park et al., 2017). Hence, we propose that the strong salt bridges may contribute to the thermostability of FiBAP, as reported previously for a hyperthermophilic PIMT, in which deletion of the interfacial residues that interact through salt bridges affected the overall thermostability of the enzyme (Tanaka et al., 2004).

Subsequently, each dimer interacts with two adjacent dimers at three contact points (boxed in Figure 4E) to form a biologically active octamer that includes two contact points (S2–S4 interface and S1–S3 interface in Figure 4C) between similar structural elements α4, α5 of one dimer, and loops T213–L220, G95–L106, and I159–L169 of the adjacent dimer. We identified additional salt bridges formed at this interface between residues R174 (α4), D178 (α4), and E214 (α5) with E171, R131, and R201, respectively (pink and purple boxes in Figure 4E). However, these corresponding residues are conserved and form salt bridges in both EcIadA and CpIadA, revealing their significance in the formation of a stable quaternary structure. In addition, the third contact point (S2–S3 interface) at the dimer-dimer interface is mainly comprised of H-bonds formed between interfacial residues (brown box in Figure 4E). These interactions at different contact points are required in BAPs for their overall stability and to maintain their high oligomeric state.

The ligand-free model included most amino acids except for residues D254–R259 and P290–L302, corresponding to the flexible loops in the molecule (Figure 4A). Intriguingly, the ligand-bound model included obvious electron density corresponding to loop P290–L302 between β14 and β15, which might have been stabilized upon ligand binding (Figure 5A). By contrast, loop D254–R259 of the Cbz-β-DL-bound FiBAP structure remained disordered. Remarkably, the Cbz-β-DL-bound FiBAP structure aligned with the ligand-free model with an RMSD of 0.433 Å, despite the good superimposition of the structural elements (Figure 5A). Such an unusually high RMSD value might be ascribed to the transformation of the missing loop (P290–L302) in the ligand-free structure into two stable antiparallel β-sheets (β14 and β15) in the ligand-bound structure (Figure 5A). Since this stabilized loop is located next to the substrate-binding site, it might function as a gate for substrate entry and/or product release (Park et al., 2017). Intriguingly, the crystal structure of FiBAP bound to Cbz-β-DL included distinct electron density only for the β-DL moiety lacking carbamazepine (Cbz) in the active site (Figure 5C).

The substrate-binding site of BAPs, referred to as the “binuclear zinc center,” is located beneath the (β/α)8 TIM-barrel motif of the catalytic domain, and comprises two Zn²⁺ (Zn1 and Zn2) ions 3.2 Å apart from each other (dotted circle in Figure 4A). Zn1 is coordinated by E156, H195, and H224 residues, while Zn2 is bound by H61, H63, E156, and D285 (Figure 5B). In the metal coordination network, Zn1 interacts with Zn2 via E156 (Figure 5B). By contrast, a carbamylated lysine residue (KCX162) in Type-II EcIadA binds two positively charged Zn ions, similar to E156 in Type-I CpIadA (Figure 5E) (Thoden et al., 2003; Park et al., 2017). In addition, E156 in FiBAP forms a cis-peptide bond with G155 in the loop between α4 and β10, which is conserved in the corresponding loop of CpIadA between G165 and E166. The cis-peptide bond favors loop bending in FiBAP, which positions E156 toward the metal, as opposed to the carbamylated lysine (KCX162) in EcIadA, suggesting that FiBAP belongs to the Type-I IadAs.

There is no significant perturbation in the coordination of the binuclear zinc center in the Cbz-β-DL-bound FiBAP structure (Figure 5C). However, structural comparison of the residues around the ligand-binding site reveals subtle changes of ~1 Å in the side-chain conformations of some residues (G68, E70, Y130, R163, R227, and S289) between Cbz-β-DL-free and -bound forms (Figure 5D). Residues such as Y130, G68, and S289 are placed closer to the ligand than those in the ligand-free structure, thereby assisting coordination and stabilization, while R227 and R163 move away to provide space for ligand binding. Notably, the role of the conserved Y130, which interacts with the O06 atom of the substrate, has been studied, and the Y137F mutation in EcIadA (corresponding to Y130 in FiBAP) reduced the rate of catalysis by three orders of magnitude (Marti-Arbona et al., 2005b). Accordingly, the conserved Y130 in FiBAP not only contributes to stabilization of the bound substrate, but also acts as a Lewis acid with the phenolic hydroxyl group of its side chain during hydrolysis of the peptide bond adjacent to Asp of the dipeptide (Marti-Arbona et al., 2005a; Park et al., 2017). The other surrounding residues that help to stabilize the bound Cbz-β-DL are the amino groups of main-chain residues S289 and G68, while T99 interacts with O04 and O05 atoms of the side-chain of the Asp residue of β-aspartyl leucine to stabilize the N-terminus of the dipeptide during substrate recognition. Additionally, R163 and R227 interact with O01 and O03 atoms and stabilize the C-terminal end of the peptide. When the substrate is recognized and stabilized, other residues such as E156, Y130, and E70 partake in acid-base catalysis during cleavage of the peptide bond via zinc metal ions Zn1 and Zn2 (Figure 5D). As stated above, all residues involved in the recognition and catalysis of a substrate are highly conserved among homologs (Figure 5E). Intriguingly, the ligand-bound structures of EcIadA and CpIadA were only obtained following mutation of conserved residues involved in metal coordination and catalysis (D285N and E80Q, respectively). By contrast, the structure of Cbz-β-DL-bound FiBAP was determined directly from wild-type FiBAP. This observation may indicate that the crystallization conditions (temperature <20°C and pH < pH 5) could have inactivated the thermostable FiBAP enzyme, favoring formation of the substrate (β-DL)-bound complex structure without being cleaved, as seen in the electron density map (Figure 5C).

To validate whether FiBAP can cleave β-DL in vivo, we constructed a Leu auxotrophic E. coli BL21 (DE3) strain by deleting the leuB gene encoding 3-isopropylmalate dehydrogenase and the iadA gene encoding BAP in E. coli (Figure 6A). We presumed that the hydrolysis of β-DL by FiBAP could provide Leu as a nutrient for the Leu auxotrophic E. coli mutant, thereby supporting its growth on M9 minimal medium lacking Leu (Figure 6B). As expected, the ΔleuB and ΔiadA double mutant grew in M9 medium supplemented with 20 amino acids, but did not grow in M9 medium containing 19 amino acids without Leu (Leu⁻; Figure 6D). Next, to investigate the effect of overexpression of FiBAP and EcIadA on bacterial growth, we constructed the pET28a-Ec_iadA and pET28a-Fi_BAP vectors (Figure 6C) and heterologously expressed the FiBAP and iadA genes in ΔleuB and ΔiadA double mutants of E. coli BL21 (DE3), respectively. The ΔleuB ΔiadA mutant did not grow on M9 medium supplemented with 19 amino acids (Leu⁻) in the presence of β-DL during 3 days of incubation, whereas the ΔleuB ΔiadA double mutant harboring the pET-iadA plasmid grew on M9 medium with 19 amino acids (Leu⁻) supplemented with β-DL, as described previously (Marti-Arbona et al., 2005a). Remarkably, the double mutant containing the pET-FiBAP plasmid could also grow on M9 medium with 19 amino acids (Leu⁻) in the presence of β-DL even at a suboptimal temperature for FiBAP (Figure 6D). This result strongly indicates that FiBAP can also cleave the β-DL dipeptide to facilitate Leu utilization for bacterial growth, suggesting that FiBAP activity enables E. coli to utilize β-aspartyl peptides as nutrient sources for cell growth.

Accordingly, we hypothesized that heat stress may cause accumulation of misfolded and/or aggregated polypeptides, including the formation of isoAsp residues, resulting in cellular growth retardation. If so, then overexpression of FiBAP may alleviate the bacterial growth defect. To validate this, we monitored the growth profiles of the ΔleuB ΔiadA double mutant grown on LB medium with and without FiBAP upon heat shock at 44°C. Consequently, after a 30 min heat shock treatment, the growth rate and cellular production of the ΔleuB ΔiadA mutant were decreased by 30% (based on the OD₆₀₀ value) compared with those of mutants expressing FiBAP and EcIadA, respectively (Figure 7A). In addition, when expressing BAP, the ΔleuB ΔiadA mutant exhibited a 15–40% decrease in the total protein concentrations of pellets, whereas the concentrations of soluble cytosolic proteins were increased by 20% (Figure 7B). Furthermore, differential protein pattern analysis by SDS-PAGE clearly indicated that expression of FiBAP could affect the soluble cytosolic protein profiles of the mutant E. coli strains, suggesting that abnormal proteins including isoAsp residues could be efficiently degraded by BAP activity under heat stress conditions (Figure 7C).

Finally, to investigate whether overexpression of FiBAP and EcIadA can reduce the isoAsp-containing protein aggregates within cells, we determined the concentration of S-adenosyl homocysteine (SAH) as the PIMT-mediated by-product derived from the isoAsp-containing peptides in heat-treated mutant strains. Indeed, RP-HPLC analysis of cell lysates demonstrated that the concentrations of SAH in both mutants expressing FiBAP and EcIadA were lower by 20 and 50%, respectively, than that of the ΔleuB ΔiadA E. coli mutant (LJW1) negative control (Figures 8A–C). These results indicate that the relative amount of isoAsp-containing proteins was decreased significantly in mutants overexpressing EcIadA (LJW2) or FiBAP (LJW3) compared with WT (LJW1) (Figure 8D), suggesting that BAP can hydrolyze isoAsp peptide bonds in the protein aggregates, and thereby alleviate the bacterial growth defect under heat shock conditions.