Resistance to Enediyne Antitumor Antibiotics by CalC Self-Sacrifice

See allHide authors and affiliations

Science  12 Sep 2003:
Vol. 301, Issue 5639, pp. 1537-1541
DOI: 10.1126/science.1086695


Antibiotic self-resistance mechanisms, which include drug elimination, drug modification, target modification, and drug sequestration, contribute substantially to the growing problem of antibiotic resistance among pathogenic bacteria. Enediynes are among the most potent naturally occurring antibiotics, yet the mechanism of resistance to these toxins has remained a mystery. We characterize an enediyne self-resistance protein that reveals a self-sacrificing paradigm for resistance to highly reactive antibiotics, and thus another opportunity for nonpathogenic or pathogenic bacteria to evade extremely potent small molecules.

Antibiotic production within actinomycetes represents a prominent instrument of natural self-preservation. These toxic secondary metabolites—small, highly functionalized molecules—are produced and often secreted, creating a local protective environment that is inhospitable to invading organisms. Because antibiotics often function by disrupting basic biological processes (e.g., DNA, protein, or cell-wall biosynthesis), antibiotic-producing organisms have evolved numerous mechanisms to evade their own chemical warfare (1). These self-resistance mechanisms are often shared among bacteria and have contributed substantially to the serious problem of antibiotic resistance among pathogenic bacteria (2). Within the repertoire of naturally occurring antibiotics, the enediynes such as calicheamicin γ1I (1) (Fig. 1) are examples of nature's ingenuity in both their architecture and effectiveness as DNA-cleaving agents. Since the susceptibility of some organisms to 1 is in the femtomolar range (3), how does an organism that produces such a potent molecule thrive? Recently, the locus of encoding for the calicheamicin (cal) biosynthesis was elucidated (4), which relied in part upon clones that carry a gene (calC) conferring resistance to 1 in vivo (5). The role of calicheamicin as a “chemical nuclease” has been extensively studied (6), and although calC was shown to convey intracellular 1 resistance in Escherichia coli (5), its mechanism of action has remained undefined.

Fig. 1.

Representative 10-membered nonchromoprotein enediynes (calicheamicin γ1I, 1; esperamicin A1, 2; dynemicin A, 3; namenamicin, 4; and shishijimicin A, 5) and a 1-cycloaromatized product (calicheamicin ϵ, 6) with structural similarities highlighted in red. ATS, aryltetrasaccharide (representing moieties A to E of molecule 1).

We created a wild-type calC construct within E. coli BL21(DE3), designated pJB2011–E. coli, for the efficient production and purification of the encoded CalC protein (7). Previous agarose gel-based assays revealed CalC to be involved in the inhibition of 1-dependent double-stranded DNA (dsDNA) scission, as evident by the qualitative lack of a DNA smear seen upon gel resolution (5). We recently established the more quantitative molecular break-light (MBL) assay in which a modified stem-loop oligonucleotide (or break-light), comprising a terminal 5′-fluorophore/3′-quencher pair and an enediyne nucleotide recognition sequence within the stem, can monitor enediyne-induced DNA scission in real time (8). Examination of 1-mediated DNA scission with the MBL assay revealed that 525 nM 1 effectively cleaved a 3.2 nM break-light solution with high spectrophotometric resolution (Fig. 2A). DNA cleavage was attenuated by CalC in a concentration-dependent manner, with one molar equivalent providing essentially complete inhibition of DNA scission. Although both 1 and esperamicin A1 (2) display similar DNA-cleaving kinetics in the MBL assay (8), the DNA-protective activity from CalC was not displayed toward 2-mediated scission (Fig. 2B).

Fig. 2.

CalC-dependent inhibition of DNA strand scission analyzed by MBL assay (20 mM Tris, pH 7.5; 3.2 nM break-light A). (A) Calicheamicin γ11 (525 nM) incubated with CalC concentrations 0.0 (open squares), 81 (solid squares), 162 (open triangles), 243 (solid triangles), 310 (open circles), and 387 (solid circles) nM. (B) Esperamicin A1 (525 nM) incubated with CalC concentrations 0.0 (solid circles) and 387 (open circles) nM. Reactions were initiated with 1 μL of 100 mM DT T at time = 0. Experiments were performed in triplicate with standard error bars positioned every 50 s for clarity.

Considering that 1 and 2 share an identical methyl trisulfide trigger for enediyne activation, the inability of CalC to protect DNA from 2-mediated cleavage suggests that CalC does not prevent the general activation of enediynes per se or specifically rescue DNA from cleaving; rather, enediyne 1 and CalC must share a specific structure-function relationship. When we extended this analysis in vivo (fig. S1), pJB2011–E. coli showed no cross-resistance toward 2 or dynemicin A (3), a structurally distinct enediyne with an anthraquinone moiety, yet it could thrive in the presence of the recently discovered 10-membered enediynes namenamicin (4) and shishijimicin A (5) (9, 10). Despite the great structural homology that 1, 2, 4, and 5 share, only species 1, 4, and 5, which share an identical and unhindered enediyne domain or “warhead,” were acted upon by CalC.

Among nine-membered enediynes (e.g., neocarzinostatin), there exist enediyne binding proteins (chromoproteins) that serve to stabilize the enediyne and possibly aid in self-protection (11), yet all conventional experiments designed to detect a parallel CalC-1 binding event have failed (12). However, monitoring the CalC-1 interaction in the presence of the reductive activator dithiothreitol (DTT), through CalC tryptophan fluorescence, revealed a sharp increase in fluorescence and suggested an interaction between activated 1 and CalC (12). Further analysis with SDS–polyacrylamide gel electrophoresis (SDS-PAGE) revealed that the reductively activated CalC-1 interaction leads to the specific proteolysis of CalC into two distinct peptides (Fig. 3B, peptides A and B). Consistent with a specific CalC-1 relationship, activated 1 failed to elicit proteolysis of other standard proteins, such as bovine serum albumin or histone 1A, and no CalC proteolysis was observed in the absence of reductive activation. From a competitive analysis of the DNA versus protein cleavage efficiencies of 1, CalC is effective at preventing DNA cleavage by sacrificing itself (Fig. 3, A and B). Subsequent high-performance liquid chromatography (HPLC) and mass spectrometry of the reaction mixtures also confirmed the proportional emergence of 6, supporting a mechanism by which activated 1 is “quenched,” possibly through direct hydrogen abstraction from CalC in a manner consistent with its action on DNA.

Fig. 3.

(A) MBL assay (20 mM Tris, pH 7.5; 3.2 nM break-light A containing 1.05 μM CalC) results from increasing titrations of 1 over a 10-min time-base scan. Titrations were performed in a range from 0.25 to 2.25 μM 1 at 0.125 μM increments, and the reactions were initiated with 1 μL of 1000 mM DT T at time = 0. Standard error bars are omitted for clarity. (B) Coomassie-stained 10 to 20% SDS-PAGE gradient gel of samples from (A). The far left lane contains the 1.05 μM CalC blank (no enediyne added). (C) In vitro profiling of CalC Gly113 mutant–dependant inhibition of DNA scission by calicheamicin. MBL assay (20 mM Tris, pH 7.5; 3.2 nM break-light A containing 1.05 μM 1) results are shown with one molar equivalent of CalC constructs wild-type CalC (pJB2011, solid circles), G113A (open circles), G113C (solid triangles), G113S (open triangles), G113Y (solid squares), and G113V (open squares), and with no CalC (diamonds). Reactions were initiated with 1 μL of 1000 mM DT T at time = 0, with standard error bars positioned every 50 s for clarity.

Considerable effort has been directed toward developing small molecules for cleaving proteins at specific amino acid residues as a tool for protein mapping (1316). Research by Jones et al. (1719), directed specifically at the application of enediyne-based models toward chemical proteolysis, has recently led to the synthesis of two enediyne-peptide hybrids that are capable of cleaving bovine serum albumin and histone H1 at specific residues (19). Although appreciable cleavage with these designed models required ∼12 hours of photoactivation and an enediyne-to-protein ratio of 10:1, this demonstration provides a precedent for the CalC cleavage mechanism—much like the classical physical-organic cycloaromatization experiments first formed the basis from which to propose the mechanism of action of naturally occurring enediynes (6, 20). N-terminal peptide sequencing and high-resolution mass spectroscopy (MS) of the CalC-1-derived peptide fragments A and B (Fig. 3B) revealed that peptide A begins at the N terminus of intact CalC, whereas peptide B begins with Arg114. In conjunction with previously demonstrated enediyne-based Gly–α carbon (Cα)–hydrogen abstraction from peptides (18), this site-specific cleavage presents an attractive model for CalC proteolysis through Gly113 backbone hydrogen abstraction. To test this postulation, five CalC Gly113 replacement mutants were constructed: G113A, G113S, G113C, G113Y, and G113V, designed to vary the Cα-hydrogen accessibility at position 113 (21). Of this set, G113A, G113S, and G113C mutants weakly tolerated the presence of 1, whereas the G113Y and G113V strains were incapable of in vivo 1 resistance (fig. S2). The in vitro activity of the mutants and the extent of proteolysis (fig. S3) were consistent with this trend (Fig. 3C and table S1), with G113A and G113S affording the best protection, albeit within approximate 8- and 14.5-fold reduction in efficacy from that of wild-type Gly113, respectively. Cumulatively, these studies further establish Gly113 as a crucial handle within the CalC resistance mechanism.

If 1 detoxification were to proceed with hydrogen abstraction from Gly113, it would leave the protein-backbone Cα radical 7 (Fig. 4B). This species, as in enediyne-based DNA scission, could readily react with oxygen under aerobic conditions to form peroxyl radical 8, which upon loss of the peroxyl radical, would lead to imine 9, the hydrolysis of which could provide peptide 10 (22). Alternatively, the decomposition of radical 8 could present the alkoxyl radical species 11, the β-scission of which would lead to carboxyamide 12 (22). Consistent with this model, liquid chromatography–mass spectroscopy (LC-MS) analysis of the CalC-1 reaction mixture detected the presence of both species 10 and 12 within the peptide fragment A, and removal of oxygen inhibited CalC cleavage (23). In addition, the LC-MS analysis correlates the disappearance of 1 with the emergence of 6, 10, and 12, suggesting direct CalC hydrogen abstraction as the mechanism of 1 inactivation. Direct hydrogen abstraction was confirmed with per-([2,2-2H2]Gly)-CalC (24). LC-MS analyses in which 1 was reacted with per-([2,2-2H2]Gly)-CalC in H2O- or D2O-based buffers gave 6 with an additional mass of +1 and +2 atomic mass units H2O and D2O buffers, respectively, consistent with direct hydrogen abstraction from Gly113 Cα and a solvent water molecule (Fig. 4A, e and f) (25).

Fig 4.

The competing mechanisms of (A) 1-based DNA scission and (B) CalC detoxification of 1. (A) Calicheamicin (1) is (a) reductively activated by a cellular thiol (RSH), forming a RS-S-1 conjugate that facilitates hetero-Michael addition and subsequently leads to (b) cycloaromatization of the warhead. In the absence of CalC, the resulting 1,4-benzyl diradical (c) abstracts hydrogen from both the sense and antisense strands of dsDNA, forming inert calicheamicin ϵ (6) and DNA-backbone radicals (d), which under aerobic conditions leads to double-stranded scission. (B) In the presence of CalC, (e) the transient 1,4-diradical species abstracts a hydrogen from both CalC Gly113 Cα and (f) a water molecule, leading to inert 6. Under aerobic conditions, the corresponding CalC Gly113 α-radical species (7) can react (g) with molecular oxygen to provide the peroxyl-radical species 8, which may lead to (h) imine 9 or (i) alkoxy 11, en route to products 10 and 12, respectively. In the case of 12, decarboxylation of the C-terminal fragment leads to a peptide beginning with N-terminal Arg114.

Previous enediyne activation studies suggest the initiating sulfide exchange chemistry is considerably slower than the rate of cycloaromatization (2629), and thus, the activating thiol (RSH) attacking the methyl trisulfide trigger can remain conjugated to 1 via a disulfide bridge (RS-S-1) (Fig. 4A, a). During this relatively long period, when the hydrophobicity and electrostatics of 1 change, it has been proposed that RS-S-1 dissociates from the DNA minor groove until hetero-Michael addition is complete. This may also provide a window of opportunity for CalC to capture and detoxify “activated warheads” before their attack on DNA. In conjunction with a tightly sequestered, enediyne biosynthesis-exportation system, CalC may then act as a cellular failsafe within the 1-producing Micromonospora, readily available to terminate any reactive enediyne species that escapes biosynthesis and transport. The CalC mechanism presents a self-sacrificing paradigm for antibiotic resistance, in which detoxification is accomplished at the expense of both the metabolite and the resistance protein. Although this is the first known example, perhaps such a self-sacrificial resistance mechanism may be more generally reserved for metabolites of extreme potency, such as the enediyne antitumor antibiotics.

Supporting Online Material

Materials and Methods

Fig. S1 to S3

Table S1

References and Notes

References and Notes

View Abstract

Navigate This Article