Resistance-Guided Mining of Bacterial Genotoxins Defines a Family of DNA Glycosylases

Abstract

Unique DNA repair enzymes that provide self-resistance against therapeutically important, genotoxic natural products have been discovered in bacterial biosynthetic gene clusters (BGCs). Among these, the DNA glycosylase AlkZ is essential for azinomycin B production and belongs to the HTH_42 superfamily of uncharacterized proteins. Despite their widespread existence in antibiotic producers and pathogens, the roles of these proteins in production of other natural products are unknown. Here, we determine the evolutionary relationship and genomic distribution of all HTH_42 proteins from Streptomyces and use a resistance-based genome mining approach to identify homologs associated with known and uncharacterized BGCs. We find that AlkZ-like (AZL) proteins constitute one distinct HTH_42 subfamily and are highly enriched in BGCs and variable in sequence, suggesting each has evolved to protect against a specific secondary metabolite. As a validation of the approach, we show that the AZL protein, HedH4, associated with biosynthesis of the alkylating agent hedamycin, excises hedamycin-DNA adducts with exquisite specificity and provides resistance to the natural product in cells. We also identify a second, phylogenetically and functionally distinct subfamily whose proteins are never associated with BGCs, are highly conserved with respect to sequence and genomic neighborhood, and repair DNA lesions not associated with a particular natural product. This work delineates two related families of DNA repair enzymes-one specific for complex alkyl-DNA lesions and involved in self-resistance to antimicrobials and the other likely involved in protection against an array of genotoxins-and provides a framework for targeted discovery of new genotoxic compounds with therapeutic potential. IMPORTANCE Bacteria are rich sources of secondary metabolites that include DNA-damaging genotoxins with antitumor/antibiotic properties. Although Streptomyces produce a diverse number of therapeutic genotoxins, efforts toward targeted discovery of biosynthetic gene clusters (BGCs) producing DNA-damaging agents is lacking. Moreover, work on toxin-resistance genes has lagged behind our understanding of those involved in natural product synthesis. Here, we identified over 70 uncharacterized BGCs producing potentially novel genotoxins through resistance-based genome mining using the azinomycin B-resistance DNA glycosylase AlkZ. We validate our analysis by characterizing the enzymatic activity and cellular resistance of one AlkZ ortholog in the BGC of hedamycin, a potent DNA alkylating agent. Moreover, we uncover a second, phylogenetically distinct family of proteins related to Escherichia coli YcaQ, a DNA glycosylase capable of unhooking interstrand DNA cross-links, which differs from the AlkZ-like family in sequence, genomic location, proximity to BGCs, and substrate specificity. This work defines two families of DNA glycosylase for specialized repair of complex genotoxic natural products and generalized repair of a broad range of alkyl-DNA adducts and provides a framework for targeted discovery of new compounds with therapeutic potential.

Keywords: AlkZ; DNA cross-link; DNA glycosylase; DNA repair; HTH_42; Streptomyces; biosynthetic gene cluster; genotoxin; intecalator; natural product; phylogenetic tree; secondary metabolism; self-resistance.

FIG 1

Phylogenetic organization of YQL/AZL proteins in Streptomyces. (A) Azinomycin B reacts with opposite strands of DNA to form an ICL, which is unhooked by AlkZ. (B) Structure of a nitrogen mustard ICL derived from mechlorethamine and unhooked by E. coli YcaQ. (C) Phylogenetic tree of YcaQ-like (YQL, blue) and AlkZ-like (AZL, red/orange; AZL2, gray) Streptomyces proteins (n = 897). The red and orange AZL clades distinguish HΦQ and QΦQ catalytic motifs. E. coli YcaQ and S. sahachiroi AlkZ proteins are labeled. (D) Sequence logos for the catalytic motifs in YQL, AZL, and AZL2 proteins. Catalytic residues are marked with asterisks. Colors correspond to side chain chemistry. (E) Copy number frequency per Streptomyces genome as a percentage of the total species analyzed (n = 436 species, 897 sequences). One-way ANOVA significance (P) values of copy number variance are 0.0078 (YQL-AZL), 0.0033 (AZL-AZL2), and 0.3305 (YQL-AZL2), the latter of which is not significant. (F) YQL/AZL coincidence frequency. The blue-shaded section represents species containing both subfamilies; the tan-shaded section represents species containing either YQL or AZL.

FIG 2

Streptomyces AZL proteins are found in diverse uncharacterized biosynthetic gene clusters. (A) Schematic depicting the workflow for identification of HTH_42 homologs in uncharacterized Streptomyces BGCs. Homologs were identified through the presence of the catalytic motif (red text in sequence alignment). The amino acid numbering is in relation to S. sahachiroi AlkZ. The corresponding Streptomyces genomes were input into antiSMASH, from which genomic distances between YQL/AZL and the nearest BGC as well as homologous clusters were extracted. (B) Violin plot showing the distribution of distances of YQL (n = 167) and AZL (n = 154) genes to the nearest BGC (in kbp; ±100 kb). The dotted line at 0 kb represents the 5′ (+)/3′ (−) termini of the nearest BGC. Thick and thin dashed lines within the plot represent the median and upper/lower quartiles, respectively. The chi-square significance (P) value between YQL and AZL data is less than 0.0001. (C) Frequency of various types of BGCs in which AZL genes were found (n = 68 clusters identified). The y axis denotes the natural product/scaffold type to which that cluster is most homologous. Black bars represent known DNA alkylators or DNA interacting metabolites, and hashed bars represent potential DNA-damaging metabolites. Lowercase letters to the right of the bars correspond to structures shown in panel D. (D) Representative compounds corresponding to BGC types in panel C. Potential reactive sites are colored red. LL-D4919α1 and hedamycin structures are shown in Fig. 3. (E and F) Nearest-neighbor analysis of AZL (E) and YQL (F). (E) Nearest genes to AZL proteins found inside and outside clusters, shown as the ratio of GO terms inside and outside and grouped by function (blue, metabolic; green, cell signaling and function; orange, genome maintenance). (F) Representative example from Streptomyces griseoviridis of nearest neighbor analysis for YQL proteins. Genes are colored according to function as in panel E (gray, unknown/hypothetical gene). These genes are invariant for all YQL proteins, with the exception of the outermost genes, in which only one instance of variance was observed.

FIG 3

AZL proteins found in characterized Streptomyces biosynthetic gene clusters. (A and B) Gene diagrams for AZL-containing BGCs producing DNA alkylating agents (A) and compounds not known to alkylate DNA (B). Gene names are labeled below the cluster diagrams. The biosynthetic scaffold produced by specific genes in the cluster is shaded gray and labeled above the respective genes. NRPS, nonribosomal peptide synthetase; PKS1/PKS2, type 1/2 polyketide synthase; (PKS), PKS-like. Chemical structures of the metabolites produced by each cluster are shown at the bottom of each panel.

FIG 4

HedH4 excises hedamycin-guanine adducts from DNA and provides cellular resistance to hedamycin toxicity. (A) HED modification of deoxyguanosine in DNA forms a HED-DNA adduct that is hydrolyzed by HedH4 to generate an abasic (AP) site in the DNA and free HED-guanine. The reactions within the dashed line are not catalyzed by HedH4. The AP nucleotide is susceptible to base-catalyzed nicking to form shorter DNA products containing either a 3′-phospho-α,β-unsaturated aldehyde (PUA; β-elimination) or a 3′-phosphate (δ-elimination). The asterisk denotes the original 5′-end of the DNA. (B) Denaturing PAGE of 5′-Cy5-labeled HED-DNA substrate and β- and δ-elimination products after treatment with enzyme or buffer (mock) for 1 h, followed by NaOH to nick the AP site. The HED-DNA reaction only goes to ∼50% completion under our reaction conditions, as shown by the two bands of equal intensity in the mock reaction. (C) HPLC-MS analysis of HED (blue) and the HED-guanine excision product from reaction of HedH4 and HED-DNA (red). Axis represents elution time (x–axis) versus relative abundance from total ion count (y–axis). Insets show mass spectra of each elution peak. (D) Wild-type and mutant HedH4 glycosylase activity for HED-DNA. Spontaneous depurination from a no-enzyme reaction (mock) is shown as a negative control. Data are means ± standard deviations (SD) (n = 3). Curves were fit to a single exponential. Representative data are shown in Fig. S3C. (E) Denaturing PAGE of HED-DNA adducts after 1 h of incubation with either buffer (mock) or bacterial alkylpurine-DNA glycosylases. (F) Denaturing PAGE of 1-h reaction products of E. coli YcaQ and HedH4 with 7mG-DNA (left) and S. bottropensis TxnU4 and HedH4 with TXNA-DNA (right). (G) Structure of NM₈-ICL. (H) Denaturing PAGE of AZB-ICL unhooking by S. sahachiroi AlkZ and HedH4 (left) and NM₈-ICL unhooking by E. coli YcaQ and HedH4 (right). Reactions were treated with buffer (mock) or enzyme for 1 h, followed by alkaline hydrolysis. MA, monoadduct. (I) HED inhibition of E. coli K-12 transformed with hedH4/pSF-OXB1 (constitutively expressed) or empty vector pSF-OXB1. The lag time is defined as the time elapsed before cells start to grow exponentially. Data are means ± SD (n = 3). Growth curves are shown in Fig. S3F and G. Significance values were determined by unpaired t test of the mean lag time values (*, 0.05 ≤ P ≤ 0.01; ***, 0.001 ≤ P ≤ 0.0001). (J) Colony dilution assay for E. coli strains with or without HedH4 exposed to increasing concentrations of HED for 1 h. Surviving fraction (%) is relative to untreated cells. Values are means ± SD (n = 3). Significance values were determined by unpaired t test of the mean sensitivity values (*, 0.05 ≤ P ≤ 0.01; **, 0.01 ≤ P ≤ 0.001).

FIG 5

YQL proteins from Actinobacteria hydrolyze simple N7-alkylguanosine lesions and interstrand cross-links. (A and B) Denaturing PAGE of reaction products of E. coli YcaQ (Eco) and YQL proteins from Thermomonospora curvata (Tcu) and Thermobifida fusca (Tfu) with 7mG-DNA (A) and NM₈-ICL (B) after 5 min and 1 h. Lane 1 of each gel is a no-enzyme control.