A single DNA point mutation leads to the formation of a cysteine-tyrosine crosslink in the cysteine dioxygenase from Bacillus subtilis

Abstract

Cysteine dioxygenase (CDO) is a non-heme iron-containing enzyme that catalyzes the oxidation of cysteine (Cys) to cysteine sulfinic acid (CSA). Crystal structures of eukaryotic CDOs revealed the presence of an unusual crosslink between the sulfur of a cysteine residue (C93 in Mus musculus CDO, MmCDO) and a carbon atom adjacent to the phenyl group of a tyrosine residue (Y157). Formation of this crosslink occurs over time as a byproduct of catalysis and increases the catalytic efficiency of CDO by at least 10-fold. Interestingly, in bacterial CDOs, the residue corresponding to C93 is replaced by a highly conserved glycine (G82 in Bacillus subtilis CDO, BsCDO), which precludes the formation of a C–Y crosslink in these enzymes; yet bacterial CDOs achieve turnover rates paralleling those of fully crosslinked eukaryotic CDOs. In the present study, we prepared the G82C variant of BsCDO to determine if a single DNA point mutation could lead to C–Y crosslink formation in this enzyme. We used gel electrophoresis, peptide mass spectrometry, electron paramagnetic resonance spectroscopy, and kinetic assays to characterize this variant alongside the natively crosslinked wild type (WT) MmCDO and the natively non-crosslinked WT BsCDO. Collectively, our results provide compelling evidence that the G82C BsCDO variant is indeed capable of C–Y crosslink formation. Our kinetic studies indicate that G82C BsCDO has a reduced catalytic efficiency compared to WT BsCDO and that activity increases as the ratio of crosslinked to non-crosslinked enzyme increases. Finally, by carrying out a bioinformatic analysis of the CDO family, we were able to identify a large number of putatively crosslinked bacterial CDOs, the majority of which are from gram-negative pathogenic bacteria.

Graphical Abstract

Introduction

Cysteine dioxygenase (CDO) is a member of the thiol dioxygenases, a family of non-heme iron-containing cupin enzymes that oxidize thiols to their corresponding sulfinates.¹ This family additionally includes cysteamine dioxygenase (ADO), 3-mercaptopropionic acid dioxygenase (MDO), mercaptosuccinate dioxygenase (MSDO),¹ and plant cysteine dioxygenase (PCO).^2,3 CDO catalyzes the oxidation of cysteine (Cys) to cysteine sulfinic acid (CSA) via incorporation of molecular oxygen.⁴ Cys is an essential protein building block as well as a precursor for the biosynthesis of the metabolites coenzyme A, glutathione, taurine, and pyruvate.⁵ However, excess free Cys is cytotoxic⁶ and neurotoxic,⁷ is easily oxidized to the insoluble cystine,⁸ and has been linked to several autoimmune and neurodegenerative diseases.^9–12 Cys oxidation is a key enzymatic step in the sulfur metabolic pathway and therefore plays a vital role in maintaining homeostasis of sulfur-containing compounds.¹

The active site of CDO contains several conserved residues that coordinate the iron atom and promote the binding and activation of Cys and O₂. X-ray crystal structures revealed a three-histidine (3-His) coordination of the iron ion instead of the 2-His, 1-carboxylate facial triad typical of non-heme iron proteins.¹³ In the resting state of CDO, the Fe(II) ion resides in a 6-coordinate pseudo-octahedral coordination environment comprised of the three His residues and three water molecules.¹⁴ Structures of CDO crystallized in the presence of Cys revealed that substrate displaces all three active site water molecules to coordinate to the iron in a bidentate fashion via its sulfur and amine groups. The sixth, open coordination site is then available for O₂ to bind end-on to the iron, facilitated by a local hydrophobic pocket.¹⁵

Eukaryotic CDO additionally contains a unique thioether crosslink between the sulfur of C93 and an ortho carbon of Y157 (Mus musculus CDO, MmCDO numbering).¹³ A similar crosslink has been observed by X-ray crystallography in only a small number of other enzymes, including galactose oxidase and the bacterial sulfite reductase NirA.^16,17 Iron, molecular oxygen, and substrate Cys must all be present for the CDO crosslink to form, and ~50% of the protein population forms this crosslink after about 800 catalytic turnovers.¹⁸ Crosslink formation increases the catalytic efficiency of the enzyme by at least 10-fold. The crosslinked “mature” and non-crosslinked “immature” forms of eukaryotic CDO travel as two distinct bands on an SDS-PAGE gel.¹⁸ The exact role of the crosslink in CDO catalysis is not fully understood. X-ray crystal structures of crosslinked wild type (WT) and the non-crosslinked C93G MmCDO are essentially identical, indicating that the crosslink does not serve to stabilize an otherwise unfavorable protein conformation.¹⁹ Also, the C93G MmCDO variant, which is incapable of C–Y crosslink formation, has comparable activity to the crosslinked WT enzyme.¹⁹ This led to the hypothesis that the crosslink serves not to directly improve enzyme efficiency, but rather to prevent deleterious effects associated with the presence of an untethered Cys residue in the enzyme active site.¹⁹

Using a combination of spectroscopic and computational techniques, we identified H155 in MmCDO as an important residue in maintaining proper positioning of the C93–Y157 crosslink. Residue H155 is a member of the S153-H155-Y157 catalytic triad motif¹³ (Figure 1) that is highly conserved among CDOs and crucial for enzymatic activity, as evidenced by the fact that the H155A MmCDO variant is ~100-fold less active than the WT enzyme.²⁰ Exposure of as-isolated H155A MmCDO (which contained similar amounts of Fe(II) and Fe(III)) to Cys led to the appearance of magnetic circular dichroism (MCD) features similar to those observed for Cys-bound WT CDO; however, the features of the Cys- and selenocysteine (Sec)-bound H155A Fe(II)MmCDO fractions were markedly blue-shifted from their WT counterparts.²¹ These shifts, in conjunction with results obtained from quantum mechanics/molecular mechanics (QM/MM) computations, suggested that in H155A MmCDO a six-coordinate (H₂O/Cys)-Fe(II) complex is stabilized by ~39 kcal/mol over its five-coordinate, Cys-only bound analogue, while in WT Cys-Fe(II)MmCDO the Fe(II) center favors a five-coordinate ligand environment.²¹ The stabilization of a six-coordinate (H₂O/Cys)-Fe(II) complex in H155A MmCDO was attributed to the increased conformational freedom of the C93–Y157 crosslink in the absence of H155, thus allowing the C93 sulfur atom to reposition itself so as to accept a hydrogen bond from a coordinated water molecule. Because in non-crosslinked eukaryotic WT CDOs the C93 residue is granted even more conformational freedom than in the H155A variant, we proposed that the low activity of non-crosslinked MmCDO arises from the formation of a similar six-coordinate (H₂O/Cys)-Fe(II)CDO complex.²¹

Figure 1.

Overlay of the active site regions of Cys-bound MmCDO (gray, PDB ID 4JTO) and BsCDO (cyan, PDB ID 4QM9).

Selective evolutionary pressure to conserve C93 in eukaryotic CDOs is curious considering this residue decreases catalytic efficiency unless it is crosslinked to Y157. However, Dominy et al. rationalized the conservation of C93 as an additional level of CDO regulation that provides fine-tuned control of intracellular Cys concentration.²² When Cys levels increase, more “mature” enzyme is produced as crosslink forms during turnover and catalytic efficiency improves. Additionally, at high levels, Cys blocks the ubiquitination and proteasomal degradation of CDO, thus increasing the intracellular CDO concentration ~10-fold. Together, these two layers of regulation give eukaryotes the capacity to drastically increase CDO activity in response to a sudden rise in intracellular Cys levels.²²

Until quite recently, it was believed that only eukaryotes can degrade Cys to CSA. However, in 2006 Stipanuk and coworkers identified numerous bacterial enzymes that could potentially possess CDO activity.²³ Four members of this family were heterologously expressed and found to oxidize Cys to CSA.²³ Although bacterial CDOs generally have overall sequence identities of less than 30% compared with MmCDO, many of the active site residues are conserved, the most notable exception being C93 (MmCDO numbering). In bacterial CDOs, this position is occupied by a highly conserved glycine (Figure 1), which implies that these enzymes are not capable of forming a C–Y crosslink. Indeed, an X-ray crystal structure of the Bacillus subtilis CDO (BsCDO) revealed that the tyrosine residue corresponding to Y157 in mouse CDO is unmodified in the bacterial enzyme.²⁴ With this important exception, the BsCDO active site structure is almost identical to that of mammalian crosslinked CDO, even in the Cys-bound state (Figure 1). Despite the drastically decreased activity of non-crosslinked eukaryotic CDOs, bacterial CDOs achieve turnover rates paralleling those of fully crosslinked eukaryotic CDOs.²³

Alignment of known CDO sequences revealed a seemingly new category of bacterial CDOs in which the R60 residue (MmCDO numbering) is substituted by a Q. These “Gln-type” CDOs were subsequently found to display increased catalytic efficiency for the conversion of 3-mercaptopropionic acid (3-MPA) to 3-sulfinopropionate than for Cys oxidation and were thus reclassified as MDOs.²⁴ Jameson and coworkers successfully produced a G95C variant of Pseudomonas aeruginosa MDO (PaMDO) capable of forming a crosslink analogous to that found in mammalian CDOs.²⁵ X-ray crystal structures revealed that when non-crosslinked, the Cys thiol excludes Y159 from its native position. A kinetic analysis showed that the K_M remained relatively unchanged between WT and ~50% crosslinked G95C PaMDO for both the native substrate 3-mercaptopropionic acid and the non-native substrate Cys.²⁵ In contrast, the k_cat values for both substrates were significantly smaller for the G95C variant compared to WT PaMDO, even though half of the active sites contained the crosslink. The loss of activity of the non-crosslinked isoform of G95C PaMDO was attributed to the mispositioning of Y159 observed in the crystal structure, but the dramatic decrease in activity of the crosslinked fraction remained unexplained.²⁵

In the present study, we prepared the G82C variant of BsCDO to determine if a single DNA point mutation could introduce into this enzyme the ability to form a C–Y crosslink. We used gel electrophoresis, mass spectrometry, electron paramagnetic resonance (EPR) spectroscopy, and kinetic assays to characterize this variant along with the natively crosslinked WT MmCDO and the natively non-crosslinked WT BsCDO. Finally, bioinformatic tools were used to identify trends of C–Y crosslink conservation across the entire CDO family.

Materials and Methods

Recombinant Gene Expression and Protein Purification.

Gene expression and protein purification of WT BsCDO were conducted as described previously²³ with a few minor alterations. In brief, a codon-optimized cdoA gene was purchased from Integrated DNA Technologies and inserted into a pQE-30 expression vector using restriction digest with BamHI-HF and HindIII-HF and subsequent ligation. Insertion was confirmed via colony PCR and the gene sequence was verified by Sanger sequencing at the University of Wisconsin-Madison Biotechnology Center. The plasmid was then transformed into Escherichia coli Rosetta 2(DE3) cells, the cells were grown in a modified TB+G medium (12 g/L tryptone, 24 g/L yeast extract, 8 mL/L glycerol with 100 mL/L 0.17 M KH₂PO₄/0.72 M K₂HPO₄, 25 mL/L 15% aspartate, and 2 mL/L 1 M MgSO₄) at 37 °C and 250 rpm, and the cdoA gene was expressed via induction with 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at an OD₆₀₀ of 8.0. Ferrous ammonium sulfate (FAS) was added to a final concentration of 500 μM at the time of induction to increase Fe incorporation into the CDO active site. Cells grew for an additional four hours after induction. Filtered cell lysate in IMAC A buffer (20 mM Tris, 5 mM imidazole, 500 mM NaCl, pH 8.0) was applied to an immobilized metal affinity chromatography column and eluted with an increasing gradient of IMAC B (20 mM Tris, 500 mM imidazole, 500 mM NaCl, pH 8.0) buffer. Fractions containing CDO as determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) were pooled and activity of the purified enzyme was confirmed qualitatively using thin-layer chromatography as previously described.²⁶

The G82C mutation was introduced into the cdoA gene using site-directed mutagenesis with the primers 5’-GGCAGAGTATTTGTTGCGCCATGG-3’ and 5’-CCATGGCGCAACAAATACTCTGCC-3’ purchased from Integrated DNA Technologies. Correct mutagenesis was confirmed via Sanger sequencing at the University of Wisconsin-Madison Biotechnology Center and variant protein was produced as described above for WT protein. Four preparations of G82C BsCDO were completed with different media and supplements to achieve variable iron loading and thus varying degrees of crosslink formation. Preps 1 and 2 were done in the modified TB+G medium described above. Preps 3 and 4 were carried out in LB medium, with Prep 3 induced at an OD₆₀₀ of ~4 and Prep 4 induced at an OD₆₀₀ of ~0.8. Preps 2, 3, and 4 were supplemented 15 min before induction with 1,10-phenanthroline to a final concentration of 100 mM in 100 mM HCl to prevent iron binding to the enzyme, as previously done by Ellis and coworkers.²⁷ The rest of the growth and purification was performed as described above for WT BsCDO.

WT MmCDO was produced using a codon-optimized cdo1 gene in the pVP16 expression vector with the gene for an attached maltose binding protein (MBP) as a solubility tag. WT MmCDO Prep 1 was purified from Escherichia coli Rosetta 2(DE3) cells containing the pVP16 expression vector with the cdo-mbp gene grown in LB medium at 37 °C and 225 rpm. At an OD₆₀₀ of 1.0, FAS was added to a final concentration of 520 μM. At an OD₆₀₀ of 2.73, expression was induced via the addition of 10 μM IPTG and the medium was supplemented with another addition of FAS to 50 μM along with 3 g/L lactose and 2 g/L Casamino acids. Cells were left to grow overnight at 25 °C and 225 rpm. Filtered cell lysate in 25 mM HEPES, 300 mM NaCl, pH 7.9 buffer was applied to a TALON column. Fractions containing CDO-MBP as determined by SDS-PAGE were pooled. Approximately 1 mg of tobacco etch virus protease was added to the pooled fractions per 50 mg of protein to cleave MBP from CDO, and the solution was dialyzed overnight against 2 L of the HEPES/NaCl buffer. The cleaved and dialyzed protein was then again applied to a TALON column. Fractions containing purified CDO as determined by SDS-PAGE were pooled and the activity of the purified enzyme was confirmed qualitatively using thin-layer chromatography as previously described.²⁶ C93G MmCDO was produced and purified following the same procedure as was used for WT MmCDO Prep 1.

WT MmCDO Prep 2 was purified from Escherichia coli Rosetta 2(DE3) cells containing the pVP-16 cdo-mbp expression vector grown in the modified TB+G medium described above at 37 °C and 250 rpm. At an OD₆₀₀ of 8.0, expression was induced with the addition of IPTG to 1 mM and FAS was added to 500 μM. Cells were grown for an additional 4 hours after induction. Filtered cell lysate in the Tris IMAC A buffer described above was applied to a TALON column and fractions containing CDO-MBP as determined by SDS-PAGE were pooled. The rest of the purification procedure was identical to that used for WT MmCDO Prep 1 described above.

SDS-PAGE Densitometry.

Purified CDO samples were run on Criterion TGX Stain-Free Precast Gels at 200 V for 42 min. Gels were stained with Coomassie Brilliant Blue G-250 and then destained in a solution of 50%:40%:10% (v/v) methanol:H₂O:acetic acid. The gels were imaged using a photo scanner. ImageJ software was employed to quantify the intensities of the two SDS-PAGE gel bands in each sample, and percent crosslinking was calculated as the ratio of the intensities of the bottom and top bands.

Mass Spectrometry.

Gel pieces were de-stained completely in 100 mM NH₄HCO₃ in 50%:50% (v/v) MeOH:H₂O and dehydrated in 25 mM NH₄HCO₃ in 50%:50% (v/v) CH₃CN:H₂O and then again in 100% CH₃CN. The samples were dried in a Speed-Vac, reduced with 25 mM dithiothreitol in 25 mM NH₄HCO₃ at 56 °C, alkylated with 55 mM chloroacetamide in 25 mM NH₄HCO₃ in darkness at room temperature, washed once in H₂O, then dehydrated as above. Primary digestion was performed by rehydrating samples with 10 ng/μL trypsin in 25 mM NH₄HCO₃/0.01% (w/v) of ProteaseMAX^™ from Promega Corp for 3 hours at 42 °C. Secondary digestion was performed for 1 hour at 37 °C using 20 ng/μL endoproteinase AspN from Roche Diagnostics in 25 mM NH₄HCO₃. Proteolysis was terminated by acidification with 2.5% trifluoroacetic acid (TFA) to 0.3% (v/v). Degraded ProteaseMAX^™ was removed via centrifugation and the peptides were solid phase extracted on a C18 column (Pierce^™ C18 SPE tips, 10 μl bed). Peptides were eluted with acetonitrile/H₂O/TFA (70%:30%:0.1% (v/v)), dried to minimum volume, and diluted to 20 μL total volume with 0.1% formic acid.

Peptides were analyzed by nanoLC-MS/MS using the Agilent 1100 nanoflow system connected to a hybrid linear ion trap-orbitrap mass spectrometer (LTQ-Orbitrap Elite^™, Thermo Fisher Scientific) equipped with an EASY-Spray^™ electrospray source (held at constant 35 °C). 2 uL of extracted peptides were loaded onto a capillary emitter column (PepMap^® C18, 3μM, 100Å, 150 × 0.075mm, Thermo Fisher Scientific). A NanoHPLC system delivered solvents A (0.1% (v/v) formic acid) and B (99.9% (v/v) acetonitrile). Peptides were loaded at 100% A and eluted by gradually increasing percent B directly into the nano-electrospray. As peptides eluted from the HPLC-column/electrospray source, survey MS scans were acquired in the Orbitrap with a resolution of 120,000 followed by collision-induced dissociation (CID)-type MS/MS with 2.0 AMU isolation and 10 msec activation time with 35% normalized collision energy fragmentation of the 30 most intense peptides detected in the MS1 scan from 350 to 1800 m/z; redundancy was limited by dynamic exclusion. Monoisotopic precursor selection and charge state screening were enabled and +1 as well as undefined charge states were rejected.

Raw MS/MS data were searched against a user defined Bacillus subtilis amino acid sequence database (UP00001570 reference proteome, 02/19/2021 download plus CDOG82C construct of interest, 5,488 total sequences) or a Uniprot Mus musculus reference database (UP000000589, 06/16/2020 download, 63723 total sequences) plus a common lab contaminants cRAP database (117 total entries) using the in-house Mascot search engine 2.7.0 (Matrix Science) with variable methionine oxidation, asparagine and glutamine deamidation, plus fixed Cys carbamidomethylation. Peptide mass tolerance was set at 10 ppm and fragment mass at 0.6 Da. Protein annotations, significance of identification, and spectral based quantification were done with the help of the Scaffold software (version 4.11.0, Proteome Software Inc., Portland, OR). Peptide identifications were accepted if they could be established at greater than 64.0% probability to achieve a false discovery rate (FDR) of less than 1.0% by the Scaffold Local FDR algorithm. Protein identifications were accepted if they could be established at greater than 14.0% probability to achieve an FDR of less than 1.0% and contained at least 2 identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm.²⁸ Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins sharing significant peptide evidence were grouped into clusters.

Sample Preparation for Spectroscopy.

Protein samples for EPR spectroscopy were prepared in a buffer of 20 mM Tris, 5 mM imidazole, 500 mM NaCl, pH 8.0. The Fe(II) and Fe(III) contents of the protein were determined via a colorimetric assay with the iron chelator tripyridyl triazine (TPTZ) and an ε₅₉₅ of 22.1 mM⁻¹ cm⁻¹.²⁹ To increase the iron content, protein samples were incubated anaerobically with a 2-fold molar excess of FAS to protein for 30 min. Chelex 100 was added to remove unbound iron from solution. Typical iron incorporation was ~60% after reconstitution. Ammonium hexachloroiridate was added in a 3-fold molar excess over iron-loaded protein to oxidize it to the Fe(III) state. Excess oxidant was removed using either a PD-10 desalting column or buffer exchange through a 10 kDa Centricon filter. EPR samples were prepared with 0.5 mM Fe(III)-loaded protein, a 10-fold molar excess of Cys, and a 15-fold molar excess of KCN.

Protein samples for MCD spectroscopy were prepared anaerobically in a glove box. Protein in a buffer of 20 mM Tris, pH 8.0, with 5 mM imidazole and 500 mM NaCl was reconstituted with FAS as described above. Tris(2-carboxyethyl)phosphine (TCEP) was added in a 3-fold molar excess over iron-loaded protein to reduce it to the Fe(II) state. Samples were then concentrated to 1.4 mM Fe(II)-loaded protein and incubated with a 2-fold molar excess of Cys. 55% (v/v) glycerol was added as a glassing agent.

Spectroscopy.

X-band EPR data were collected using a Bruker ELEXSYS E500 spectrometer. Sample temperature was maintained at 20 K by an Oxford ESR 900 continuous flow liquid He cryostat regulated by an Oxford ITC-503S temperature controller. All EPR spectra were obtained using the following experimental parameters: frequency = 9.381 GHz; microwave power = 12.62 mW; modulation amplitude = 3 G; and modulation frequency = 100 kHz. EPR spectral fits were performed using the SIMPOW program.³⁰

Low-temperature MCD spectra were recorded with a Jasco J-715 spectropolarimeter in conjunction with an Oxford Instruments SpectromagPT 7 Tesla magnetocryostat. MCD spectra are presented as the difference between spectra obtained with the magnetic field aligned parallel and antiparallel to the light propagation axis to eliminate contributions from the circular dichroism background and glass strain.

Kinetic Assays.

Quantitative kinetic assays were performed using a combination of two methods. The first method was adapted from Stipanuk and coworkers and Jameson and coworkers and employed an Intrada column from Imtakt on a Waters ultra-high performance liquid chromatography mass spectrometer (UPLC-MS).^19,31,32 Protein samples were thawed in hand and then exchanged into a buffer of 200 mM 2-(N-morpholino)ethanesulfonic acid (MES), 150 mM NaCl, 0.1 mM bathocuproine disulfonate (BCS), pH 6.1. The concentration of Fe(II)-loaded protein was determined via the TPTZ assay described above.²⁹ 100 μL reactions of 50 μM protein and 1–50 mM cysteine were run at 37 °C for 12 min. 20 μL reaction aliquots were quenched every 3 min by the addition of 180 μL of equal volumes of 1 M HCl and 1 mM asparagine, which served as an internal standard. Quenched reaction aliquots were centrifuged at 15,000 × g for 3 min and the supernatants were transferred to a 96-well plate. 2 μL of samples were injected onto the UPLC-MS and applied to the Intrada column under 80% buffer A (acetonitrile with 0.1% (v/v) formic acid), 20% buffer B (100 mM ammonium formate) and species eluted during a gradual increase to 80% buffer B were analyzed with the mass detector. Enzyme activity was quantified using the ratio of CSA mass peak to asparagine mass peak in comparison with a calibration curve. Values for k_cat and K_M were determined using the concentration of Fe(II)-loaded enzyme and the Michaelis-Menten analysis reported by Johnson.³³

The second method was adapted from Jameson and coworkers.³⁴ Protein was exchanged into a buffer of 100 mM sodium phosphate, 20 mM NaCl, 0.1 mM BCS, pH 7.5. The concentration of Fe(II)-loaded protein was determined using the colorimetric TPTZ assay described above²⁹ and used as the concentration of active enzyme. A 0.64 mM solution of 5,5’-dithio-bis-[2-nitrobenzoic acid] (DTNB, Ellman’s reagent) in phosphate buffer was prepared and stored at 4 °C for up to a week. A 100 mM stock of Cys in phosphate buffer was prepared and adjusted to a pH of 7.5 using NaOH. For each protein sample, 5 reactions of 7 μM protein were run at 37 °C and 300 rpm. The reaction was initiated by addition of substrate Cys to protein to a final concentration of 1, 3, 5, 10, or 20 mM. Reaction aliquots of 2.5 μL were quenched every 2 min for 24 min by addition of the aliquot to a well in a 96-well plate containing 97.5 μL of 0.6 mM DTNB. Absorbance readings at 412 nm for each well were taken on a Tecan SPARK plate reader using 350 μL capacity BRAND clear flat-bottom 96 well plates. A standard curve of 0.1–20 mM Cys was prepared with each protein assay and used to calculate the concentration of Cys remaining at each reaction time point. Values for k_cat and K_M were again determined using the concentration of Fe(II)-loaded enzyme and the Michaelis-Menten analysis reported by Johnson.³³

Sequence Similarity Network and Multiple Sequence Alignment.

A sequence similarity network (SSN) of the CDO family was generated using the InterPro-defined Cysteine Dioxygenase Type I family (IPR010300) with the resources provided by the Enzyme Function Initiative-Enzyme Similarity Tool (EFI-EST).³⁵ Upon retrieval of the sequences and primary network generation, the dataset was pared down to further refine the information provided and yield a suitable file size. Only sequences with lengths between 125 and 300 amino acids were used. A 65% representative node cutoff was used to accommodate the large number of sequences and an E-value of 10⁻¹⁸ was chosen as the minimum stringency cutoff. The SSN was visualized using Cytoscape 3.9.1.³⁶ Taxonomy labels for clusters were generated using the Taxonomy Name/ID Status Report function from the NIH Taxonomy Database website to identify the complete taxonomy lineage from the taxonomy ID for each sequence in the SSN cluster.³⁷ An in-house MATLAB R2018a script was then used to sort the taxonomy lineage identifiers in frequency order.

A multiple sequence alignment (MSA) of all sequences in the InterPro Cysteine Dioxygenase Type I family was generated using the Super5 algorithm in the Multiple Sequence Comparison by Log-Expectation (MUSCLE) v5 program.³⁸ The alignment was visualized in SnapGene Viewer 4.2.11 (www.snapgene.com) and the positions that aligned with C93 and Y157 in the MmCDO sequence were identified. An in-house MATLAB R2018a script was used to identify all CDO sequences that contained analogous C and Y residues at these positions, and these sequences were labeled in the SSN as being putatively capable of crosslink formation. SnapGene was used to generate consensus sequences of subsets of CDOs using their aligned FASTA format sequences generated by MUSCLE. Residues were numbered by the MmCDO residue with which they aligned. When comparing several consensus sequences, conserved sequence gaps found in all consensus sequences were removed for clarity.

Results

Production of Crosslinked G82C BsCDO Variant.

The goal of this study was to determine if a single G-to-C substitution in BsCDO could lead to the formation of a C–Y crosslink analogous to that found in eukaryotic CDOs. For eukaryotic CDOs, a characteristic double band pattern is observed on an SDS-PAGE gel, as the crosslinked and non-crosslinked forms travel as proteins with slightly different apparent molecular weights.¹⁸ We found that while WT BsCDO produced a single SDS-PAGE band, as expected, the G82C variant traveled as two bands (Figure 2), like eukaryotic CDOs. Crosslink formation in eukaryotic CDOs is known to be a side reaction of Cys oxidation and thus dependent on the presence of enzyme-bound iron and substrates Cys and O₂.¹⁸ By varying the iron content of our growth media during the production of G82C BsCDO, we were able to change the relative intensities of the two bands on the SDS-PAGE gel (Figure 2), further suggesting that the G82C BsCDO variant is capable of crosslink formation. ImageJ software was used to quantify the intensities of the two SDS-PAGE gel bands in each sample, and the fraction of crosslinked protein was calculated as the ratio of the intensities of the bottom and top bands.³⁹ As expected, by increasing the concentration of iron in the growth medium, the fraction of crosslinked G82C BsCDO variant increased (Table 1). While the difference in percentage of crosslinked protein determined for the three low-iron G82C BsCDO preparations (Preps 2, 3, and 4) likely falls within the error margin associated with SDS-PAGE densitometry, a significantly higher fraction of crosslinked variant was obtained from the high-iron preparation (Prep 1).

Figure 2.

SDS-PAGE gels of (A) WT BsCDO, (B) G82C BsCDO Prep 1, (C) G82C BsCDO Prep 2, (D) G82C BsCDO Prep 3, and (E) G82C BsCDO Prep 4. The different conditions used for Preps 1–4 and percentage of crosslinked protein in each sample are provided in Table 1.

Table 1.

Growth conditions and percentage of crosslinked WT BsCDO and different preparations of the G82C BsCDO variant as determined via integration of the SDS-PAGE gel band intensities in Figure 2

Lane	Sample	Growth Conditions	Crosslinked Fraction
A	WT BsCDO	Modified TB medium, 500 μM FAS	0%
B	G82C BsCDO High Iron Prep 1	Modified TB medium, 500 μM FAS	62%
C	G82C BsCDO Low Iron Prep 2	Modified TB medium, 100 μM 1,10-phenanthroline	34%
D	G82C BsCDO Low Iron Prep 3	LB medium, 100 μM, 1,10-phenanthroline, IPTG induction at OD600 ≈ 4	42%
E	G82C BsCDO Low Iron Prep 4	LB medium, 100 μM 1,10-phenanthroline, IPTG induction at OD600 ≈ 0.8	38%

Interestingly, because the G82C BsCDO samples from the low iron Preps 3 and 4 were run at a lower concentration on the SDS-PAGE gel, a splitting of the lower band associated with crosslinked protein into a pair of two closely spaced bands could be discerned (Figure 2, lanes D and E). Such a splitting has not been observed for eukaryotic CDOs and may reflect two slightly different confirmations of the denatured, crosslinked forms of G82C BsCDO.

Mass Spectrometry.

To confirm that the double band pattern observed for the different preparations of G82C BsCDO on the SDS-PAGE gel in Figure 2 was indeed due to the formation of a C82–Y141 crosslink in a subset of protein monomers, MS experiments were performed. The top and bottom bands of a high iron G82C BsCDO sample (Prep 1) were excised from an SDS-PAGE gel, digested using a trypsin/AspN combination, and subjected to a MS analysis. Complete sequence coverage was achieved, which indicated with >95% certainty that the protein in both bands was G82C BsCDO. The total ion chromatograms of the protein extracted from the top and bottom gel bands were identical except for the relative intensities of the two peaks associated with the peptides containing the non-crosslinked C82 and Y141 residues (Figure 3). The drastically reduced intensities of these peaks in the sample from the lower gel band indicate that in this protein fraction the corresponding peptides are crosslinked. The newly observed middle band on the SDS-PAGE gel for G82C BsCDO was also excised and analyzed using MS. For this protein fraction, the relative intensities of the two peptide peaks were found to be comparable to, albeit slightly more intense than, those displayed by the protein extracted from the bottom gel band (Figure S2), in support of our hypothesis that the two lower gel bands reflect two slightly different confirmations of the denatured crosslinked enzyme.

Figure 3.

Relative intensity of (A) the (H)DHGQSI⁸²CCAMVLEGK(L) fragment and (B) the (R)MVSLHV¹⁴¹YSPPLE(D) fragment in the mass spectra of the protein extracted from the upper and lower SDS-PAGE gel bands of the high iron G82C BsCDO Prep 1. Analogous results were obtained with WT MmCDO (Figure S1).

Because no new species due to dipeptide formation could be observed by scanning raw data or by using crosslinking software, we performed an analogous MS experiment for WT MmCDO (Prep 2, ~65% crosslinked). The top and bottom bands were excised from an SDS-PAGE gel and processed as described above for G82C BsCDO. Again, while we observed different relative intensities of the two peaks associated with the peptides containing the non-crosslinked C and Y residues in the ion chromatograms of the protein extracted from the top and bottom gel bands (Figure S1), we were unable to identify a new peptide fragment containing the C–Y crosslink. This result is identical to what Dominy et al. observed in their original MS analysis of the top and bottom bands of WT MmCDO.¹⁸ Thus, our MS experiments provide compelling evidence that the G82C BsCDO variant is capable of C–Y crosslink formation.

EPR spectra of (Cys/CN⁻)-bound Fe(III)CDO complexes.

Previous studies revealed that the S = ½ EPR spectrum displayed by (Cys/CN⁻)-bound Fe(III)MmCDO, which mimics the putative Fe(III)-superoxo intermediate, is sensitive to the absence or presence of the C–Y crosslink.²⁰ Crosslinked (Cys/CN⁻)-bound Fe(III)MmCDO was found to exhibit a larger spread of the EPR g-values than the fully non-crosslinked C93A variant and the non-crosslinked fraction of as-isolated MmCDO.²⁰ We were able to replicate these results using as-isolated WT MmCDO (Prep 1), which in our hands was mainly crosslinked, and its C93G variant, which is unable to form this crosslink (Figure 4A and B).

Figure 4.

EPR spectra of the (Cys/CN⁻)-Fe(III) adducts of (A) WT MmCDO Prep 1, (B) C93G MmCDO, (C) WT BsCDO, and (D) high iron G82C BsCDO Prep 1. Features between ~3250 and 3400 Gauss are due to a contaminant in the instrument cavity. The dashed black vertical lines indicate the g₁ and g₃ values for the crosslinked and non-crosslinked fractions of WT MmCDO, and the dashed green vertical lines indicate the g₁ and g₃ values for the crosslinked and non-crosslinked fractions of G82C BsCDO. All EPR g values for the different species obtained from spectral fits are provided in Table 1.

As expected, the EPR spectrum of (Cys/CN⁻)-bound WT Fe(III)BsCDO displays a single, rhombic EPR signal (Figure 4C) with g values that are similar to those exhibited by the (Cys/CN⁻) adduct of the C93G MmCDO variant (Figure 4B), which is also unable to form a C–Y crosslink. Alternatively, the EPR spectrum of (Cys/CN⁻)-bound G82C BsCDO (Figure 4D) is nearly identical to that of WT MmCDO (Figure 4A), showing contributions from both crosslinked (major fraction) and non-crosslinked (minor fraction) protein. The g values obtained from fits of the EPR spectra in Figure 4 are listed in Table 2. These fits also allowed us to determine the relative contributions from the two different S = ½ signals for (Cys/CN⁻)-bound WT MmCDO and G82C BsCDO, and thus the ratio of crosslinked to non-crosslinked Fe(III)-bound protein. It is important to note that this ratio is not expected to match the ratio of crosslinked to non-crosslinked protein determined using SDS-PAGE densitometry, because the former only depends on the fraction of CDO containing (Cys/CN⁻)-bound Fe(III) centers, while the latter is determined by the complete CDO population regardless of Fe loading or oxidation state. To test the hypothesis that iron was required for crosslink formation, we additionally collected an EPR spectrum of G82C BsCDO from cells grown in a low-iron environment (Prep 4). As expected, a higher relative contribution from the (Cys/CN⁻) adduct of the non-crosslinked protein to the EPR spectrum of this sample was observed (Figure 5). The same trend was seen when comparing EPR spectra of (Cys/CN⁻)-bound WT MmCDO samples obtained using different growth and purification conditions (Figure S3). Together with our SDS-PAGE gel and MS results, these EPR data provide compelling evidence that a C–Y crosslink forms in G82C BsCDO.

Table 2.

g values and relative contributions from the crosslinked and non-crosslinked protein fractions obtained from fits of the EPR spectra shown in Figures 4, 5, and S3–S5

Sample	Species	% Contribution	g1	g2	g3
A. WT MmCDO Prep 1	Crosslinked	65	2.378	2.235	1.935
	Non-crosslinked	35	2.335	2.205	1.951
B. WT MmCDO Prep 2	Crosslinked	88	2.378	2.235	1.935
	Non-crosslinked	12	2.335	2.205	1.951
C. C93G MmCDO	Non-crosslinked	100	2.349	2.243	1.941
D. WT BsCDO	Non-crosslinked	100	2.327	2.227	1.948
E. G82C BsCDO High Iron Prep 1	Crosslinked	79	2.370	2.223	1.935
	Non-crosslinked	21	2.326	2.223	1.949
F. G82C BsCDO Low Iron Prep 4	Crosslinked	67	2.370	2.223	1.939
	Non-crosslinked	33	2.326	2.223	1.949

Figure 5.

EPR spectra of the (Cys/CN⁻)-Fe(III) adducts of (A) high iron G82C BsCDO Prep 1 and (B) low iron G82C BsCDO Prep 4. Spectra are shown as solid lines and fits as dashed lines. The ratios of crosslinked to non-crosslinked Fe(III)-bound active sites as determined via fitting of these spectra are included. Features between ~3250 and 3400 Gauss are due to a contaminant in the instrument cavity.

MCD Spectra of Cys-bound WT and G82C Fe(II)BsCDO.

In previous studies of as-isolated Cys-bound WT MmCDO, features at 32,000 and 15,700 cm⁻¹ were shown to arise from S_Cys → Fe(II) and S_Cys → Fe(III) charge transfer (CT) transitions, respectively.²⁶ Interestingly, in MCD spectra of the H155A MmCDO variant, the S_Cys → Fe(II) CT transitions were found to be markedly blue-shifted from their WT counterparts.²¹ These shifts were attributed to the formation of a six-coordinate (H₂O/Cys)-Fe(II) complex in the variant, rather than a five-coordinate Cys-only Fe(II) adduct as observed for the WT enzyme. The ~100-fold decreased catalytic efficiency displayed by the H155A MmCDO variant thus likely stems from the lack of an open coordination site for O₂ of the Cys-bound Fe(II) center. To ensure that the G82C substitution and partial crosslink formation did not cause the binding of a water molecule or any other drastic perturbation to the catalytically relevant Fe(II) form of BsCDO, we collected MCD spectra of Cys-bound WT and G82C Fe(II)BsCDO (Figure 6). Both spectra are dominated by the expected S_Cys → Fe(II) CT band at ~33,000 cm⁻¹. The close similarity of these spectra indicates that the G82C substitution and partial formation of the C–Y crosslink in BsCDO causes no major geometric or electronic structural changes of the Fe(II) center. Thus, we can confidently attribute any major activity differences between WT and G82C BsCDO to the absence or (partial) presence of the crosslink.

Figure 6.

Variable temperature MCD spectra at 7 T of Cys-bound (A) WT Fe(II)BsCDO and (B) low iron G82C Fe(II)BsCDO Prep 2.

Kinetic Assays of WT and G82C BsCDO with native substrate Cys.

We recently adopted two protocols to determine K_M and k_cat values for previously uncharacterized CDO variants. The first, based on work by Stipanuk and coworkers and Jameson and coworkers, involves the use of UPLC-MS to monitor both substrate consumption and product formation.^19,31,32 The second, pioneered by Jameson and coworkers, utilizes Ellman’s reagent to monitor substrate depletion spectrophotometrically.³⁴ To demonstrate that we successfully adopted these assays, we first incubated WT MmCDO with various concentrations of Cys and found that the enzyme displayed Michaelis-Menten kinetics with a K_M of 5.5 mM and k_cat of 0.8 s⁻¹ (Table 3), in good agreement with results reported by Stipanuk and coworkers (K_M of 4.5 mM and k_cat of 0.72 s⁻¹).³¹ We then confirmed that WT BsCDO has comparable activity to fully crosslinked eukaryotic CDO, despite the absence of the C–Y crosslink (Table 3). G82C BsCDO showed reduced catalytic efficiency compared to WT BsCDO, and activity increased as the ratio of crosslinked to non-crosslinked enzyme increased. While in Table 3 we report the percentage of crosslinked enzyme for each sample as determined using SDS-PAGE densitometry, this method accounts for all protein in the preparation, whereas kinetic assays only probe the active, Fe(II)-bound form of the enzyme. Thus, we expect a qualitative, rather than a quantitative correlation between the percentage of crosslinked total protein and enzymatic activity of G82C BsCDO. Michaelis-Menten fits of all kinetic data are provided in Figure S6.

Table 3.

Kinetic parameters of different MmCDO and BsCDO species as determined by activity assays

Sample	% Crosslinked In SDS-PAGE Gel	KM (mM)	kcat (sec−1)	kcat/KM (mM−1sec−1)
MmCDO Prep 2	Not measured	5.5 ± 0.3	0.8 ± 0.1	0.15 ± 0.03
BsCDO	0	6.6 ± 0.8	0.4 ± 0.1	0.05 ± 0.02
G82C BsCDO High Iron Prep 1	62	10.7 ± 0.3	0.17 ± 0.02	0.016 ± 0.002
G82C BsCDO Low Iron Prep 2	34	11.5 ± 0.1	0.0041 ± 0.0003	0.00036 ± 0.00003

Our least crosslinked sample of G82C BsCDO (Prep 2) showed a drastic (~150-fold) reduction in catalytic efficiency relative to WT BsCDO. This result is somewhat unexpected, as one would anticipate a linear relationship between activity and percentage crosslink formation if only the crosslinked fraction were active. This drastic reduction in activity was persistent across all low-iron preparations of G82C BsCDO (Preps 2, 3, and 4), regardless of whether or not the protein was reconstituted with iron before the activity assay (data not shown). Because we also observed very weak EPR signals for samples of the low-iron growths of G82C BsCDO after reconstitution with Fe(II) and incubation with Cys and CN⁻, and since previous studies revealed that different CDO variants incorporate different amounts of Zn(II), we hypothesize that G82C BsCDO produced under low Fe conditions or in the presence of a chelator incorporates different metals in place of Fe.⁴⁰ Reconstitution with FAS could lead to non-specifically bound Fe that yields seemingly normal Fe-loading values in the TPTZ assay. In this scenario, inactive, mis-metallated protein and non-specifically bound Fe in G82C BsCDO Prep 2 that increase the apparent G82C Fe(II)BsCDO concentration could explain the unexpectedly large difference in k_cat between the high iron Prep 1 and low iron Prep 2 of G82C BsCDO.

Discussion

Since their discovery and initial kinetic and structural characterization seventeen years ago,^23,24 no further investigations of bacterial CDOs have been reported. In the present study, we took advantage of the fact that WT BsCDO is unable to form the C–Y crosslink that is found in all eukaryotic CDOs. By substituting the glycine at position 82 with a cysteine, we were able to successfully create a variant of BsCDO capable of crosslink formation, as confirmed by SDS-PAGE gel electrophoresis, mass spectrometry, and EPR spectroscopy. MCD data indicate that the Cys-bound (catalytically relevant) form of the G82C Fe(II)BsCDO variant has essentially the same active site electronic structure as WT BsCDO. Kinetic analyses revealed that C82 in BsCDO has a similar effect on the rate of Cys oxidation as does the analogous C93 in MmCDO. Introduction of the Cys residue leads to a drop in enzyme activity, but crosslink formation between C82 and Y141 recovers, at least partially, the activity of the WT enzyme. Because the population of crosslinked G82C BsCDO correlates with the amount of iron present during protein production, it is reasonable to assume that C–Y crosslink formation in this variant also occurs as a result of an unproductive reaction with L-cysteine and dioxygen, as in the case of WT MmCDO.¹⁸

In their initial study, Stipanuk and coworkers discovered only 38 putative bacterial CDOs in sequence databases, and all appeared incapable of crosslink formation.²³ With the exponential increase in available protein sequence data over recent years, we decided to conduct an updated and more comprehensive analysis of putative CDOs across all organisms. As of July 2022, the InterPro Cysteine Dioxygenase Type I family contained 11,270 proteins, 9,159 of which were from bacteria, 1,823 from eukaryotes, and 187 from archaea.⁴¹ It is clear from these numbers alone that CDOs appear to be much more prevalent in bacteria than originally thought, and that the sulfur metabolic pathways of bacteria may not yet be fully understood. Figure 7 shows a sequence similarity network (SSN) of all proteins between 125 to 300 amino acids in length that belong to the InterPro Cysteine Dioxygenase Type I family (IPR010300).⁴¹ Nodes represent groups of proteins with ≥ 65% sequence identity and edges are drawn between pairs of nodes for which the BLAST E-values are less than a user-defined, upper limit threshold, E.

Figure 7.

SSN of the InterPro CDOI family at (A) low stringency (E=10⁻²¹) and (B) high stringency (E=10⁻²⁶). Eukaryotic CDOs are shown in green, bacterial CDOs in blue, and archaeal CDOs in red. CDOs that are putatively capable of crosslink formation based on the presence of residues analogous to both C93 and Y157 (MmCDO numbering) are shown as larger dots. Each major cluster is labeled with the taxonomic category to which the majority of CDOs in that cluster belong.

At a low stringency value (E=10⁻²¹, Figure 7A), a few large clusters appear. All eukaryotic CDOs are clustered together, along with a grouping of CDOs from the Fibrobacteres-Chlorobi-Bacteroidetes (FCB) clade of bacteria and a grouping from proteobacteria and planctomycetes bacteria, all of which appear capable of crosslink formation based on the presence of Cys and Tyr residues analogous to C93 and Y157 in MmCDO. A large cluster of CDOs from actinobacteria and a small cluster of CDOs from bacilli appear incapable of crosslink formation because they lack the Cys residue corresponding to C93 in MmCDO and instead typically possess a G at this position. A large cluster of CDOs from proteobacteria is split roughly down the middle between putatively crosslinked and non-crosslinked CDOs. Some putatively crosslinked CDOs from actinobacteria cluster with the putatively crosslinked half of CDOs from proteobacteria. Generally, clustering occurs based on superkingdom or phylum classification, while also depending on putative crosslinking ability. The majority of bacterial CDOs that appear capable of crosslink formation are from proteobacteria. Some CDOs from actinobacteria may also be capable of crosslink formation; however, they share greater sequence identity with the allegedly crosslinked proteobacterial CDOs than with the non-crosslinked actinobacterial CDOs.

At a high stringency value (E=10⁻²⁶, Figure 7B), smaller clusters emerge based on the organisms’ class or order. We observe that the FCB clade CDOs remain closest to eukaryotic CDOs. The major types of proteobacterial CDOs putatively capable of crosslink formation are from beta- and gammaproteobacteria, specifically Burkholderia, Xanthomonadales, and Vibronales. Vibronales and planctomycetes CDOs cluster with archaeal CDOs. These are all gram-negative pathogenic bacteria, suggesting the possibility that crosslinked CDO was obtained through horizontal gene transfer from a eukaryotic host. In their discovery of non-crosslinked bacterial CDOs, Dominey et al. suggested that CDO may aid spore formulation in bacteria by decreasing Cys levels and thus promoting disulfide bond formation.²³ However, most of the putatively crosslinked bacterial CDOs in our SSN are from organisms that do not form spores. Streptomyces is the largest genus of actinobacteria that appears capable of CDO crosslink formation, and these CDOs have greater sequence identity with those from Burkholderia than with other, non-crosslinked actinobacterial CDOs. Interestingly, BsCDO, which was used in this study, is not as similar to eukaryotic CDOs as some other non-crosslinked CDOs are.

To better understand the conservation of specific structurally and catalytically relevant residues across different clusters of CDOs, a multiple sequence alignment (MSA) was performed. Figure 8 shows the aligned consensus sequences of eukaryotic, putatively crosslinked bacterial, and non-crosslinked bacterial CDOs. Residues are numbered according to the MmCDO scheme and important residues are highlighted. The three Fe-binding His residues as well as the S-H-Y “catalytic triad” are conserved across all CDOs. However, several key differences can be noted. At position 60, non-crosslinked bacterial CDOs contain a Q, which is typical of MDOs, rather than the R typical of CDOs.²⁴ Interestingly, however, BsCDO maintains the R residue here. Instead of a Y at position 58, which is thought to aid in substrate binding, non-crosslinked bacterial CDOs contain a conserved R.¹⁵ Meanwhile, putatively crosslinked bacterial CDOs lack the cis-proline bond thought to position residue 164 and differentiate CDOs from MDOs.⁴²

Figure 8.

MSA comparing the consensus sequences of crosslinked eukaryotic, putatively crosslinked bacterial, and non-crosslinked bacterial CDOs. One letter amino acid codes are used, where “X” represents a position where no particular amino acid is conserved and “-” represents a gap where no amino acid exists in that consensus sequence. Residues are labeled vertically above each position with MmCDO numbering. Residues conserved across all three groups of CDOs are highlighted in green. Differences in conserved amino acid residues are highlighted in red. Legend: *, Fe(II)-binding His residues; !, S-H-Y “catalytic triad”; Y, additional substrate-binding Tyr in CDOs; #, active site R and Q residues distinguishing CDOs from MDOs and MSDOs; c, cis-Pro peptide bonds in CDOs.

Thus, while the 3-His triad and S-H-Y triad are absolutely conserved, residues known to be involved in substrate binding in thiol dioxygenase enzymes are not uniformly conserved across (putatively) crosslinked and non-crosslinked CDOs. This may indicate differences in substrate specificity across clusters of CDOs. Considering the differences between the effect on enzyme activity of installing a crosslink into BsCDO performed in this study and into PaMDO as accomplished previously by Jameson and coworkers, this may also imply that these substrate binding residues and the absence or presence of crosslink cooperatively affect enzyme turnover rate.²⁵

Conclusions

Collectively, our data show that it is possible to build a C–Y crosslink into BsCDO, further confirming previous observations that the spatial proximity of these two residues contributes substantially to their ability to crosslink.^25,43 We have demonstrated that as in eukaryotic CDO, the non-crosslinked isoform displays decreased catalytic efficiency.¹⁸ Our bioinformatic analysis of the CDO family revealed the existence of many putatively crosslinked bacterial CDOs, the majority of which are from gram-negative pathogenic bacteria. Additionally, we observed that residues involved in substrate binding are not conserved across eukaryotic, putatively crosslinked bacterial, and non-crosslinked bacterial CDOs, indicating potential relationships between substrate specificity, C–Y crosslinking, and enzyme activity.