S. rimosus ATCC 10970 was obtained from China General Microbiological Culture Collection Center (CGMCC 4.1438). E. coli JM109 and BL21(DE3) were used for general cloning and protein expression, respectively. LB medium was used for E. coli cultivation. Kanamycin was used at a final concentration of 50 μg/mL. Restriction enzymes and Q5 DNA polymerase were purchased from New England Biolabs (United States). DNA manipulations, competent cell preparation, and transformation were performed as described previously (Sambrook and Russell, 2001). Hydrogen peroxide (30%) solution was purchased from Sigma-Aldrich. The 96-well UV-transparent microplates were purchased from Corning.

The plasmids and primers used in this study are listed in Supplementary Tables 1, 2, respectively. The five catalase genes of S. rimosus ATCC 10970 were amplified through PCR with their corresponding primer pairs (Supplementary Table 2) using the genomic DNA of S. rimosus as a template. The plasmid pET28a was digested by NdeI and XhoI, and then joined with each of the five catalase gene fragments by Gibson assembly (Gibson et al., 2009) to generate the five overexpression plasmids (Supplementary Table 1). All the plasmids were verified by DNA sequencing and transformed into E. coli BL21(DE3) to obtain the five overexpression strains.

The strains were cultivated in LB supplemented with 50 μg/mL of kanamycin at 37°C and 220 rpm until an OD600 of 0.6 was reached, at which time final concentrations of 100 μM isopropyl-β-D-thiogalactoside (IPTG), 0.5 mM 5-aminolevulinic acid, and 0.2 mM FeCl₃ were added. The cell cultures were further cultivated for 8 h at 28°C and 180 rpm. The five N-terminal His₆-tagged catalases (Cat1-5) were purified by Nickel-NTA affinity chromatography according to the previously reported protocol (Fan et al., 2012), concentrated by centrifugation, and exchanged to ultrafiltration buffer (50 mM Na₂HPO₄-NaH₂PO₄, 10% (v:v) glycerol, pH 7.4) using 10-kDa Amicon Ultra tubes (Centriplus, Merck Millipore). The purities of the proteins were evaluated by 10% SDS-PAGE. The protein concentrations were determined by Bradford assay using bovine serum albumin (BSA) as a standard. The proteins were stored at –70°C.

The UV-vis spectra of five catalases were collected on a Beckman Coulter DU800 spectrophotometer using a cell with an optical path of 1 cm. The heme concentrations were calculated using absorption at 406 nm (ε₄₀₆ = 102 mM^–1 cm^–1) (Brill and Sandberg, 1968). The heme occupancy of each catalase was the ratio of heme concentration to the corresponding protein concentration (by Bradford assay).

Circular dichroism (CD) spectra of five catalases were monitored in the far-UV range (195–260 nm) in 50 mM Na₂HPO₄-NaH₂PO₄ (pH 7.4), at 30°C on a Chirascan^TM CD Spectrometer (Applied Photophysics, United Kingdom) using a cell with an optical path of 1 mm. The stability of the five catalases were monitored by CD signals at 222 nm (representing the content of α-helices) from 20 to 90°C (raising 1°C at a time, and holding for 5 s before data collection) in 50 mM Na₂HPO₄-NaH₂PO₄ (pH 7.4).

The activities of the five catalases were measured using the spectrophotometric method described previously (Beers and Sizer, 1952). A typical 200 μL reaction mixture consisted of 16 mM H₂O₂, purified catalase (1.06-10.32 nM), and 50 mM Na₂HPO₄-NaH₂PO₄ (pH 7.4). The absorbance at 240 nm (representing the concentration of H₂O₂) was monitored continuously in 96-well plates at 30°C using an EnSpire^TM multimode Plate Reader (PerkinElmer Inc., United States). The reaction rates were calculated using the absorbance changes at 240 nm within the first 1 min, and averaged for three independent reaction samples. One unit of enzyme activity is defined as the amount of activity required for the conversion of 1 μmol of H₂O₂ into water and oxygen per minute at 30°C (Zeng et al., 2011).

The catalase activities under different pH conditions were measured at 30°C. A series of buffers were used, including 50 mM NaAc-HAc (pH 4.0 – 6.0), 50 mM NaH₂PO₄-Na₂HPO₄ (pH 6.0 – 8.0), and 50 mM Na₂CO₃-NaHCO₃ (pH 8.0 - 10.0). The reaction rates were calculated using the absorbance changes at 240 nm within the first 1 min. The concentrations of the enzymes used in the assay were Cat1: 1.06 nM, Cat2: 1.35 nM, Cat3: 3.83 nM, Cat4: 1.11 nM, and Cat5: 10.32 nM.

The catalase activities at different temperatures (30 to 65°C) were measured in 50 mM NaH₂PO₄-Na₂HPO₄ (pH 7.4). The reaction rates were calculated using the absorbance changes at 240 nm within the first 1 min. The concentrations of the enzymes used in the assay were Cat1: 1.06 nM, Cat2: 1.35 nM, Cat3: 3.83 nM, Cat4: 1.11 nM, and Cat5: 10.32 nM.

The apparent K_m and k_cat values of the five catalases were determined by non-linear least squares fitting using reaction rates with different H₂O₂ concentrations (7.5 - 50 mM) at 30°C in 50 mM NaH₂PO₄-Na₂HPO₄ (pH 7.4). The specific activities of the five catalases were determined using the reaction rates when 16 mM H₂O₂ was used as the substrate. The reaction rates were calculated using the absorbance changes at 240 nm within the first 1 min. The concentrations of the holoenzymes (deduced from the detected heme occupancy of each purified catalase) used in the reactions were Cat1: 1.15 nM, Cat2: 1.17 nM, Cat3: 1.26 nM, Cat4: 1.05 nM, and Cat5: 6.05 nM.

The catalase homologs were searched using standalone BLASTP program (Camacho et al., 2009) with the 21 known catalases reported previously as query sequences (Supplementary Table 3; Zamocky et al., 2008). Pfam search was performed using HMMER 3.3 (Eddy, 2011) with reported Pfam motifs (typical catalase: PF00199, catalase-peroxidase: PF00141, Mn-catalase: PF05067) (Zamocky et al., 2008; El-Gebali et al., 2019). Sequence comparison was performed using EMBOSS 6.5.0 (Rice et al., 2000). Those homologs with Pfam scores lower than the trusted cutoffs of corresponding Pfam motifs were removed, along with those with shorter lengths (typical catalase clade 1: <450 residues, clade 2: <600 residues, clade 3: <450 residues, catalase-peroxidase: <600 residues, Mn-catalase: <180 residues). The phylogenetic trees were constructed by a neighbor-joining method using MEGA X with the default parameters (Kumar et al., 2018). The information on all catalase homologs, including their GenBank accession numbers and sequences, are provided as Supplementary Dataset 1; the phylogenic trees of three families of catalases are provided as Newick files in Supplementary Dataset 2.

To explore the abundance and diversity of catalases in bacteria, we carried out a virtual survey of 13,360 completed bacterial genomes in the NCBI Genomes database (as of Sep 2020) using the 21 well studied catalases covering all three families (Supplementary Table 3) as query sequences. As the numbers of sequenced genomes varied dramatically among different taxa, no more than 10 genomes from each genus were randomly selected as representatives for further analysis. As a result, a total of 2,634 genomes were selected, in which 4,931 catalase homologs were identified. The homologs were further filtered by removing 71 proteins with Pfam scores lower than the trusted cutoffs or with abnormal protein sequence lengths (Supplementary Figure 1). Pairwise sequence comparisons of the final 4,860 catalase homologs clearly showed the three distinct protein families consisting of the typical catalases, catalase-peroxidases, and Mn-catalases (Figure 1A).

We next subjected the proteins of each catalase family to phylogenetic analysis, which revealed diverse groups within each family (Figure 1). In the phylogenetic tree, 2,615 typical catalase homologs fell into seven groups (group 1-7, Figure 1B). The proteins of group 5 have apparently larger sizes ranging from 600 to 850 residues. They showed relatively lower sequence similarity to the homologs from other groups (Supplementary Figure 2A). The proteins from groups 6 and 7 have 450-550 residues. The group 6 catalase homologs are mainly from Actinobacteria, while a small portion of them are from Proteobacteria and Firmicutes. However, the majority of group 7 proteins are from Proteobacteria, with a small minority of them coming from Bacteroidetes and Firmicutes. The proteins of groups 2, 3, and 4 are predominantly 500-600 residues in size. Among them, the group 4 catalase homologs are from Proteobacteria, while the proteins of groups 2 and 3 come from various phyla (Figure 1B). The proteins from group 1 are shorter (450-500 residues), and display relatively lower sequence identity to the homologs from other groups of this family (Supplementary Figure 2A). The proteins of this group are mainly from Firmicutes.

Phylogenetic analysis of the 1,681 catalase-peroxidase homologs appeared to form four major groups (Figure 1C). However, these groups are closely related, as they are all approximately 700-850 residues in size, and share high sequence identity (Figure 1C and Supplementary Figures 1B, 2B), forming a tight catalase-peroxidase protein family. Among them, the proteins in group 3 are mainly from Proteobacteria, and a minority of them come from Bacteroidetes and Firmicutes. The proteins in group 4 are from Actinobacteria, Proteobacteria, and Bacteroidetes. The proteins in group 1 come from various phyla, while group 2 proteins are mainly from Proteobacteria.

For the Mn-catalase family, the 564 homologs clearly formed ten groups, and their inter-group sequence identities were typically lower than 40% (Figure 1D and Supplementary Figure 2C). The proteins from groups 1 and 2 are significantly shorter in length (group 1: 150-200 residues, group 2: 200-250 residues), and these proteins are from Firmicutes. Most proteins of group 3 and 4 have medium lengths (250-300 residues), and are mainly from Proteobacteria and Firmicutes, respectively. In contrast, the proteins of groups 5 and 6 are larger (300-350 residues) than others, and they are shown to be from Actinobacteria and Proteobacteria, respectively (Figure 1D). The proteins of group 7 differ in length and taxa distribution. The group 8 protein members are mainly from Carnobacterium and Lactobacillus, while those of group 9 are from Proteobacteria, Bacteroidetes, and Deinococcus. The proteins of group 10 vary over a broad range in length, and their host strains include Actinobacteria, Firmicutes, and Cyanobacteria.

We first surveyed and compared the richness and diversity of catalases among different bacteria at the phylum level. The distribution patterns of the three catalase families varied among different phyla (Figure 2). Overall, the typical catalases are widely distributed except in Chlorobi, and the catalase-peroxidases are present in many phyla, while Mn-catalases mainly exist in Deinococcus-Thermus and Firmicutes. The phyla, such as Acidobacteria, Actinobacteria, Bacteroidetes, Planctomycetes, and Proteobacteria, are rich in typical catalases and catalase-peroxidases but lack Mn-catalases. In contrast, the phylum of Deinococcus-Thermus is rich in Mn-catalases and typical catalases, but not catalase-peroxidases. The majority of catalases in Chlorobi are catalase-peroxidases, while the typical catalases are the ones dominant in Spirochetes. Cyanobacteria and Chloroflexi have relatively fewer catalases than other phyla. The shortage of typical catalases in Cyanobacteria has been also observed previously (Zamocky et al., 2008; Johnson and Hug, 2019).

We then further analyzed the distributions of three catalase families in 81 well-sampled genera (each genus with ten selected genomes). A notable high frequency of typical catalases was observed in soil bacteria Streptomyces, Massilia, Rhodococcus, and the pathogenic bacteria Pseudomonas (Figure 2). Indeed, all the selected genera under the phylum of Actinobacteria are high in typical catalases (mainly groups 5 and 6), and majority of them are also rich in catalase-peroxidase (mainly group 4) (Figure 2). Some genera of Actinobacteria also contain Mn-catalases, such as Amycolatopsis. The Bacillus and Paenibacillus strains are rich in various groups of typical catalases and Mn-catalases, while their close relatives, the thermophilic Geobacillus strains, mainly contain catalase-peroxidases and Mn-catalases. The Staphylococcus strains are rich in typical catalases (groups 5 and 7) but lack of catalase-peroxidases and Mn-catalases. The facultative or strict anaerobes under the Firmicutes phylum, including Enterococcus, Lactobacillus, and Clostridium, are also short of catalase-peroxidases but rich in typical catalases and Mn-catalases. The representative opportunistic pathogenic bacteria, such as Enterobacter, Klebsiella, and Proteus strains, possess all three catalase families, including the typical catalases (mainly group 5), catalase-peroxidases (group 4), and Mn-catalases (group 3). However, the richness in all the three catalase families is not observed in other pathogenic bacteria like Yersinia and Pseudomonas strains.

In particular, we were interested to know the abundance and diversity of catalases in Streptomyces strains, which have complex life cycles and are the prolific producers of natural products with therapeutic applications (Bentley et al., 2002; Flärdh and Buttner, 2009; Musiol-Kroll et al., 2019). To get an overarching view of the distribution of catalases in Streptomyces, we extracted all the catalase homologs in 271 annotated Streptomyces genomes available in NCBI Genomes database, resulting in 811 typical catalase homologs and 268 catalase-peroxidase homologs, with no Mn-catalase homologs identified. In other words, there are 2.99 typical catalases and 0.99 catalase-peroxidases on average in each Streptomyces genome. Specifically, the catalase homologs in four well-characterized antibiotic-producing Streptomyces strains, including S. coelicolor A3(2), S. venezuelae ATCC 10712 (ISP5230), S. avermitilis MA-4680, and S. rimosus ATCC 10970, are listed in Supplementary Table 4. It is worth noting that, although an average of 0.99 catalase-peroxidase homolog per genome was found in Streptomyces, it was clear that some Streptomyces strains contain no catalase-peroxidase (such as S. avermitilis MA-4680 and S. rimosus ATCC 10970), whereas some other strains can have more than one homolog (like S. venezuelae ATCC 10712; Supplementary Table 4).

Remarkably, sequence analyses revealed that all the identified 268 catalase-peroxidase homologs belong to group 4 (Figure 3). For the 811 typical catalase homologs, the majority of them belong to group 6 (71%), with the remaining homologs falling into groups 5 (19%), 3 (8%), and 7 (3%). This distribution pattern was consistent with that of the ten randomly selected Streptomyces genomes (Figure 2). As shown in Figure 3, the group 5 and 6 typical catalases in Streptomyces may be further divided into two subgroups based on sequence analysis.

The above bioinformatics analysis revealed the richness of typical catalase homologs in Streptomyces. To verify the predicted catalase function of these homologs, we selected and characterized all five typical catalase homologs (Cat1-5) identified in S. rimosus ATCC 10970, the oxytetracycline producing strain (Zhang et al., 2006). The five proteins with N-terminal His₆-tags were overproduced in E. coli BL21(DE3), and purified to homogeneity (Supplementary Figure 3). The purified proteins were reddish brown in solution, and their UV-vis spectra showed characteristic Soret bands at 405 nm (Supplementary Figure 4), supporting that they were heme-binding enzymes. All five enzymes could catalyze the decomposition of H₂O₂, as the absorbance at 240 nm of the reaction mixtures decreased continuously for all five enzymes while such an absorbance change was not observed for the reaction without enzymes (Supplementary Figure 5). These results confirmed that all five proteins are catalases. The specific activities of these catalases were determined, in which the Cat5 (group 5, Clade 2) showed relatively low activity (Table 1). Overall, the specific activities of these typical catalases were comparable to those of reported catalases (Switala and Loewen, 2002; Sooch et al., 2014).

We then further analyzed the properties of these catalases by evaluating the optimal reaction pH and temperature for each of them. The Cat1, Cat3, and Cat4, all belonging to group 6, showed optimal pH at 7.4, and their activities decreased significantly when pH changed (Figure 4). Cat2 (group 3) showed high activities under alkaline pHs, with an optimal pH of 9.0. The Cat1-4 showed highest activities at 35°C, and their activities decreased significantly above 35°C. Surprisingly, Cat5 (group 5) showed nearly constant activity over a broad pH range of 5.0-10.0, and remained active at temperatures from 30 to 65°C (Figure 4). The thermostabilities of five catalases were also characterized by circular dichroism (CD) spectroscopic analysis. All five catalases showed negative peaks at 208 and 222 nm (Supplementary Figure 6), which corresponded to the α-helices of the conserved α + β catalase fold (Nicholls et al., 2000; Chelikani et al., 2004; Díaz et al., 2012). The thermal denaturation processes of five catalases were monitored by the CD signals at 222 nm (Supplementary Figure 6). Among them, Cat1 (Tm: 52°C) and Cat3 (Tm: 53°C) were less resistant to heat, while Cat4 (Tm: 75°C) displayed moderate heat tolerance. Surprisingly, although Cat2 showed complete loss of activity at 65°C, no significant change of the CD signal at 222 nm was observed up to 80°C. These results indicated that, under the above assay conditions, the tertiary structure of Cat2 was likely damaged, but most of the α-helix secondary structures remained unchanged. Cat5 displayed excellent thermostability up to 80°C, which explained why this enzyme remained catalytically active at 65°C as shown in Figure 4.

Due to the two-stage reaction mechanism, catalases did not follow Michaelis-Menten kinetics except at low H₂O₂ concentrations (Switala and Loewen, 2002; Zamocky et al., 2008). Therefore, the apparent K_m and k_cat values for the five catalase reactions were determined at low substrate concentrations (not more than 50 mM) (Table 1 and Supplementary Figure 7). Among them, Cat5 displayed lower catalytic efficiency. Overall, the kinetic values were comparable to the reported data of known typical catalases (Switala and Loewen, 2002; Jia et al., 2016, 2017).