Designer enediynes generate DNA breaks, interstrand
cross-links, or both, with concomitant changes in the
regulation of DNA damage responses
Daniel R. Kennedy*, Jianhua Ju†, Ben Shen†‡§, and Terry A. Beerman*¶
*Department of Pharmacology and Therapeutics, Roswell Park Cancer Institute, Buffalo, NY 14263; and †Division of Pharmaceutical Sciences, ‡University of
Wisconsin National Cooperative Drug Discovery Group, and §Department of Chemistry, University of Wisconsin, Madison, WI 53705
Communicated by Thomas C. Bruice, University of California, Santa Barbara, CA, August 31, 2007 (received for review June 20, 2007)
The ability of the radiomimetic anticancer enediyne C-1027 to
induce ataxia-telangiectasia mutated (ATM) and ATM and Rad3related (ATR)-independent damage responses was discovered to
reside in its unique ability to concurrently generate robust
amounts of double-strand breaks (DSBs) and interstrand cross-links
(ICLs) in cellular DNA. Furthermore, a single substitution to the
chromophore’s benzoxazolinate moiety shifted DNA damage to
primarily ICLs and an ATR- but not ATM-dependent damage response. In contrast, single substitutions of the chromophore’s
␤-amino acid component shifted DNA damage to primarily DSBs,
consistent with its induction of conventional ATM-dependent
damage responses of the type generated by ionizing radiation and
other radiomimetics. Thus, phosphatidylinositol 3-kinase-like protein kinase regulation of DNA damage responses is dictated by the
relative proportions of DSBs and ICLs.
DNA double-strand break 兩 C-1027 兩 radioimetic 兩 ATM 兩 ATR
C
ells use several members of phosphatidylinositol 3-kinaselike protein kinases (PIKKs), including ataxia-telangiectasia
mutated (ATM) and ATM and Rad3-related (ATR), to initiate
cell cycle checkpoint responses to DNA damage (1). The induction of double-strand breaks (DSBs) into cellular DNA rapidly
triggers ATM activation, which promotes a kinase cascade
involving phosphorylation of the signal transducers Chk1 and
Chk2 and downstream targets such as p53, all of which contribute to cell cycle arrest (1, 2). Similar to ATM, ATR activates
Chk1, p53, and, to a lesser degree, Chk2 to initiate cell cycle
checkpoints in response to lesions such as interstrand cross-links
(ICLs), which stall DNA replication (2).
Although ATM’s role in responding to DSBs was primarily
based on ionizing radiation (IR) studies, radiomimetic enediynes
such as neocarzinostatin, C-1027, and calicheamicin also have
provided insight (3). Enediynes are a structurally diverse group
of compounds whose chromophores bind to the DNA minor
groove and subsequently undergo Bergman cycloaromatization,
generating two free radicals (4, 5). These radicals can induce
DSBs when hydrogen atoms in close proximity, but on opposite
DNA strands (6), are abstracted from deoxyribose.
Cellular responses to radiomimetic treatment are similar to
IR, which include ATM-dependent activation of p53-Ser-15 and
Chk2-Thr-68 (7, 8) and enhanced cell death in cells lacking ATM
(5). That responses to DSBs are ATM-dependent regardless of
the damaging agent has led to the general conclusion that cells
require ATM to activate DNA damage responses to DSBs (5, 6).
The sole exception is C-1027, where cells deficient in either ATM
or ATR phosphorylate p53-Ser-15 and Chk2-Thr-68 as readily as
wild type and are diminished only when both PIKKs are absent
(9, 10). Furthermore, C-1027 is similarly toxic to wild-type,
ATM-, or ATR-deficient cells, because cell death hypersensitivity occurs only in the absence of both (10).
Remarkably, minor modifications of the C-1027 chromophore
can alter the PIKK dependence of the DNA damage response
(11). Two analogs containing single modifications of the
17632–17637 兩 PNAS 兩 November 6, 2007 兩 vol. 104 兩 no. 45
␤-amino acid moiety of the C-1027 chromophore, 20⬘-deschloroC-1027 (deschloro) and 22⬘-deshydroxy-C-1027 (deshydroxy)
(Fig. 1), induced IR-like ATM-dependent damage responses
(11). In contrast, a single modification of the benzoxazolinate
moiety created 7⬙-desmethyl-C-1027 (desmethyl) (Fig. 1), which,
like C-1027, induced ATM-independent DNA damage responses
(11). Although desmethyl, like all enediynes, readily induced
DSBs under cell-free conditions, few DNA breaks were detected
in cells. Nevertheless, like C-1027 and its other analogs, this
agent was extremely cytotoxic and activated cell cycle checkpoint
proteins (11).
That C-1027 appears to contradict the PIKK paradigm of
ATM’s essential role in the response to DSBs is puzzling, and the
shifting cellular DNA damage responses to structurally similar
C-1027 analogs adds to the quandary. One possible explanation
stems from the observation that C-1027 can induce additional
types of DNA damage, such as ICLs under cell-free anaerobic
conditions (12). If such lesions also were induced into cells,
perhaps they contribute to the decision of which PIKK regulates
the damage response. This study examines the capacity of C-1027
and its analogs to induce DSBs and ICLs under both cell-free and
cellular conditions. Whether the lesions induced relate to ATM
and/or ATR activation of cellular damage response pathways was
revealed.
Results
C-1027 and Its Analogs Can Induce ICLs Under Cell-Free Conditions.
Under aerobic conditions, all C-1027 analogs induced DSBs.
However, deschloro, deshydroxy, and desmethyl are 4⫻, 30⫻,
and 50⫻ less potent, respectively, than C-1027 (11). To examine
whether the analogs also induced ICLs under anaerobic cell-free
conditions (12), drug-treated, linearized plasmid DNA was used
to evaluate the amount of double-stranded DNA (dsDNA) that
remained after alkaline denaturation.储 At 13 nM C-1027, ⬇13%
dsDNA remained, which increased to ⬎30% by 130 nM (Fig. 2
A and B). Desmethyl induced ICLs as potently as C-1027 (Fig.
2 A and B). Deschloro induced limited ICLs that could be
detected consistently only at higher concentrations (4,000 nM),
whereas deshydroxy-induced ICLs were not readily detected
(Fig. 2 A and B). In Table 1, analog induction of DSBs and ICLs
were normalized to the C-1027 concentration, which produced
20% linear DNA by topological forms conversion under aerobic
Author contributions: D.R.K. and T.A.B. designed research; D.R.K. performed research; J.J.
and B.S. contributed new reagents/analytic tools; D.R.K. and T.A.B. analyzed data; and
D.R.K., B.S., and T.A.B. wrote the paper.
The authors declare no conflict of interest.
Abbreviations: ATM, ataxia-telangiectasia mutated; ATR, ATM and Rad3-related; DSB,
double-strand break; ICL, interstrand cross-link; IR, ionizing radiation; PIKK, phosphatidylinositol 3-kinase-like protein kinase; SV40, simian virus 40.
¶To
whom correspondence
roswellpark.edu.
储Under
should
be
addressed.
E-mail:
Terry.Beerman@
the denaturation conditions used, an average of ⬍4% of the dsDNA remained.
© 2007 by The National Academy of Sciences of the USA
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0708274104
conditions and 10% dsDNA remaining after denaturation, respectively. Interestingly, desmethyl, the least potent analog at
inducing DSBs, was potent at inducing ICLs, whereas deschloro
was ⬎300-fold less potent (Table 1). The relative potencies for
inducing aerobic DSBs was C-1027 ⬎ deschloro ⬎ deshydroxy ⬎
desmethyl, whereas for anaerobic ICLs the relative potency was
C-1027 ⫽ desmethyl ⬎ deschloro (Table 1).
C-1027 and Desmethyl Induce ICLs into Intracellular Simian Virus 40
(SV40) DNA. Although the extent of enediyne induction of cell-
free and cellular DSBs is generally well correlated (13), there is
Fig. 2. C-1027 and desmethyl induce ICLs under cell-free anaerobic conditions. (A) Induction of ICLs produced by C-1027 or analogs. Linearized pBR322
plasmid DNA was incubated with the indicated drug for 4 h at room temperature under anaerobic conditions and examined as described in Materials and
Methods. (Upper) Representative gel of C-1027 and desmethyl. (Lower) Representative gel of deschloro and deshydroxy. ND, nondenatured control. (B)
Quantitation of ICLs induced by C-1027 and analogs. ICL induction is based on
the relative percentage of dsDNA compared with a nondenatured control.
Kennedy et al.
no comparable information on cellular ICLs because no such
activity has been reported. Thus, we compared the ability of
C-1027 and its analogs** to induce both DSBs and ICLs on an
intracellular SV40 DNA target (14, 15).
As shown previously (14), 8 nM C-1027 induces significant
amounts of linear intracellular SV40 DNA (20%), which increases in a concentration-dependent manner to ⬇40% (Fig.
3A). Deschloro also induced DSBs in a concentration-dependent
manner, although, consistent with previous studies of genomic
DNA damage, 8-fold higher drug concentrations in comparison
to C-1027 were required (⬇60 nM) (Fig. 3A) (11). Desmethyl
initially induced a minimal amount of intracellular SV40 DSBs,
but additional breaks were not observed even at concentrations
as high as 10 ␮M, which is also consistent with previous genomic
DNA studies (Fig. 3A) (11). The potency of the C-1027 family
members at inducing intracellular SV40 DSBs is C-1027 ⬎
deschloro ⬎⬎ desmethyl.
The detection of intracellular SV40 ICLs was based on the
ability of isolated and subsequently linearized viral DNA to
remain double stranded after alkaline treatment. For C-1027,
ICLs were observed at 40 nM and significantly increased at 200
nM (Fig. 3B). Thus, C-1027 induces ICLs not only under cell-free
anaerobic conditions, but also in cells. However, at the concentrations where ICLs are observed, the accompanying cleavage of
full-length linear SV40 DNA would tend to underestimate ICLs
(Fig. 3).
Desmethyl also induced intracellular SV40 ICLs (Fig. 3B). At
200 nM, desmethyl appears to induce ICLs somewhat more
potently than C-1027. However, exact comparisons are difficult
because desmethyl’s limited intracellular strand scission activity
does not decrease the population of full-length linear SV40 DNA
(Fig. 3). In contrast, deschloro-induced intracellular ICLs were
not readily detected (Fig. 3B).
**Because deshydroxy-induced ICLs were not detectable in the cell-free studies, they were
not tested in this assay.
PNAS 兩 November 6, 2007 兩 vol. 104 兩 no. 45 兩 17633
BIOCHEMISTRY
Fig. 1. Structures of the enediyne chromophores of C-1027 and its engineered analogs. (A) Structures of the C-1027 chromophore and its biochemical subunits:
benzoxazolinate, deoxy aminosugar, ␤-amino acid, and enediyne core. (B) Structures of the C-1027 analogs desmethyl, deschloro, and deshydroxy. The rectangles
indicate the composition and location of the C-1027 chromophore modifications.
Table 1. Comparison of drug-induced DSBs vs. ICLs
Variable
C-1027
Desmethyl
Deschloro
Deshydroxy
DSB induction
relative to C-1027
ICL induction
relative to C-1027
1
0.02
0.25
0.033
1
1
⬍0.003
Not detectable
Drug potency for inducing DSBs and ICLs was normalized to C-1027 and
is shown in the second and third columns, respectively.
C-1027 and Desmethyl Induce ICLs into Intracellular Genomic DNA.
The SV40 data provide evidence that enediynes can induce
intracellular ICLs (Fig. 3B). We next determined the presence of
ICLs in genomic DNA by using a modification of the alkaline
single-cell gel electrophoresis or comet assay. The amount of
strand breaks induced by treatment of cells with 50 or 20 Gy IR
(Fig. 4 A and B respectively) is indicated by the comet tail
observed after electrophoresis under denaturing conditions.
However, pretreatment before IR with an agent that induces
ICLs but not strand breaks would decrease the IR-induced
comet tail because the cross-links render fragmented DNA
resistant to alkaline-induced DNA strand separation (16). Thus,
a reduction in comet score relative to IR alone is representative
of the amount of ICLs.
As expected with a strand scission agent, 10 nM deschloro
induced large comet tails, whereas deschloro ⫹ IR treatment
resulted in larger tails than either IR or deschloro treatment
alone (Fig. 4A). Thus, the deschloro ⫹ IR score of 3.9 is
consistent with the intracellular SV40 studies because deschloro
induces DNA strand breaks with no evidence of ICLs in genomic
DNA (Fig. 4C). For desmethyl, only minimal tail lengths were
detected even at a 100 nM concentration, but the desmethyl ⫹
IR score was reduced with increasing desmethyl levels from 3.2
to 1.6, consistent with an agent inducing ICLs (Fig. 4 A and C).
For comparison, treatment with the ICL-inducing drug, bizelesin, reduced the IR score to 0.67 (Fig. 4C).
Fig. 3. C-1027 or analogs can induce DSBs and/or ICLs in intracellular SV40
DNA. (A) Quantitation of DSBs produced by C-1027 and analogs. SV40infected BSC-1 cells were treated with the indicated drug for 4 h at 37°C, and
SV40 DNA was isolated as described in Materials and Methods and examined
for topological forms conversion. The percentage of full-length linear DNA
represents the drug potency at inducing intracellular SV40 DSBs. (B) DNA from
control and drug-treated SV40-infected BSC-1 cells was isolated, converted to
a full-length linear form with BamH1, and examined as described in Materials
and Methods. Induction of ICLs, as in Fig. 2B, was based on full-length linear
DNA that remained after alkaline denaturation. ND, nondenatured control.
17634 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0708274104
Fig. 4. Desmethyl and C-1027 induce ICLs in genomic DNA. (A) Representative alkaline comets assessing ICLs in HCT116 cells treated with deschloro and
desmethyl for 4 h and then mock-treated or exposed to IR as indicated.
Induction of ICLs was based on the reduction in IR comet tails by using the
procedure described in Materials and Methods. (B) The comet analysis for
deschloro and C-1027 is essentially as described in A. (C) Comets from A were
scored as described in Materials and Methods, and the black dotted line
represents the expected contribution of IR-induced DNA breaks to the comet
signal. Bizelesin, an ICL drug, is included as a positive control. (D) The comet
tails for deschloro and C-1027 were scored as described in C.
In the SV40 studies, C-1027 induced both DSBs and ICLs in
intact cells (Fig. 3). Thus, detection of C-1027-induced ICLs by
comet may be limited because the strand breaks may mask their
detection. A 20-Gy dose of IR was used to induce a moderate
comet tail (score of 2.1) to enhance detection of either increases
or decreases in comet size (Fig. 4B). C-1027 at 0.3 nM also
induced a moderate comet signal, but the C-1027 ⫹ IR combined
comet score of 1.7 was below the 2.1 of IR alone (Fig. 4D).
Similarly, 1 nM C-1027 further enhanced the comet tail, and yet
again the C-1027 ⫹ IR tail score of 1.9 resembled IR alone (Fig.
4 B and D). In contrast, deschloro ⫹ IR again gave a larger comet
signal (3.1) than either deschloro or IR alone, consistent with the
absence of ICLs (Fig. 4 B and D).
The Requirement of ATR to Induce Cellular DNA Damage Responses
Varies Among C-1027 and Its Analogs. The paradigm that ATM
responds to DSBs whereas ATR responds to ICLs has never been
applied to a compound like C-1027, which simultaneously induces both of these lesions. C-1027’s ATM- and ATRindependent DNA damage responses, which are diminished only
Kennedy et al.
Discussion
The cellular response to C-1027-induced DNA damage was the
first demonstrated exception of ATM’s essential role in responding to IR or a radiomimetic (9, 10). It was considered unlikely
that the nature of C-1027-induced strand breaks relates to its
ATM independence because they are similar to those of IR and
other radiomimetics (5, 18). Our discovery that C-1027 and
desmethyl induced ICLs under both cell-free and cellular conditions would be consistent with their ATM-independent activation of DNA damage response proteins (11). Similarly, deschloro’s and deshydroxy’s induction of DSBs, but not ICLs, is in
agreement with their activation of ATM-dependent damage
Kennedy et al.
Fig. 5. Cells require ATR to activate DNA damage response proteins and
avoid hypersensitive growth inhibition in response to desmethyl. HCT116 cells
were pretreated for 48 h with mock (siRNA buffer) (ATR ⫹) or siATR (ATR ⫺)
before treatment with equicytotoxic concentrations of the indicated drug for
1 h at 37°C (A and B) or 3 days (C). (A and B) Cellular extracts were analyzed by
Western blotting and probed with an antibody specific for phosphorylated
Chk1-Ser-345 (A) or with an antibody specific for phosphorylated Chk2-Thr-68
or p53-Ser-15 (B). (C) Growth inhibition IC50 was calculated based on the
concentration of drug required to reduce cell growth by 50%. The fold
differences in the IC50 values values were used to determine the ratio of
growth inhibition between mock siRNA- or siATR-treated cells.
responses (11). Thus, the consequences of minor modifications
to the enediyne chromophore of C-1027 shift not only its ability
to induce DSBs and/or ICLs, but also change PIKK regulation
of the cellular damage responses (11).
For C-1027, concurrent induction of DSBs and ICLs provides
an unprecedented opportunity to test the paradigm that cells
require ATM to activate DNA damage responses to strand
breaks, whereas ATR is required to respond to cross-links. First,
C-1027’s ATM-independent activation of proteins such as p53Ser-15 and Chk2-Thr-68 is consistent with the presence of ICLs,
which would trigger their activation by an ATR cascade (Figs. 4
and 5). Second, cells with diminished ATR levels fully activate
p53-Ser-15 and Chk2-Thr-68 after C-1027 treatment, which is
consistent with DSBs triggering activation of the ATM kinase
cascade (10). Finally, the ability to induce both types of lesions
explains why only cells deficient in both of these PIKKs exhibit
reduced activation of DNA damage responses proteins and
hypersensitive growth inhibition (10).
Strikingly, removal of either the 20⬘-chloro or 22⬘-hydroxy
group on the ␤-amino acid moiety of the chromophore converts
C-1027 to a conventional radiomimetic enediyne (i.e., an agent
that only induces DSBs and solely ATM-dependent DNA damage responses) (11). In contrast, conversion of the 7⬙-methoxy
group to a hydroxyl on the benzoxazolinate component retains
only the strong cellular ICL activity of the parent, which was
surprising considering it is a radiomimetic compound under
cell-free conditions (Figs. 2–4). However, this finding explains
desmethyl’s ability to induce cellular cytotoxicity and G2 cell
cycle arrest without inducing robust amounts of DSBs (11).
PNAS 兩 November 6, 2007 兩 vol. 104 兩 no. 45 兩 17635
BIOCHEMISTRY
in the absence of both PIKKs (10), can now be explained by
C-1027’s unique DNA-damaging capabilities. Similarly, the finding that deschloro and deshydroxy induce robust levels of DSBs
with few or no ICLs explains why they revert to a classical
ATM-dependent damage response (11). Likewise, desmethyl’s
robust ICL-inducing activity is consistent with its ATMindependent induction of DNA damage responses. If our concept is correct, loss of ATR activity in the presence of ATM
should primarily impact DNA damage responses induced by
desmethyl because, unlike C-1027 and deschloro, it induces few
cellular DSBs.
To examine the role of ATR in response to C-1027 family
members, cells were treated with either siRNA targeted against
ATR (siATR), which reduced ATR levels by 85% (data not
shown), or a mock (buffer control) (10). The phosphorylation of
Chk1-Ser-345, an immediate downstream target of the ATRinduced kinase cascade (17), would be expected to be reduced
by all of the drugs, but for desmethyl, we anticipated a more
extreme reduction. Its 2-fold decrease after treatment of ATRdeficient cells with 1 nM C-1027 is consistent with previous
studies (Fig. 5A) (10). In response to 3 nM deschloro, ATRdeficient cells also displayed an ⬇2-fold decrease in Chk1
phosphorylation (Fig. 5A). In contrast, after 60 nM desmethyl
treatment, Chk1 phosphorylation was almost completely repressed (Fig. 5A).
Next, drug-induced phosphorylation of Chk2-Thr-68, an upstream general transducer kinase, and p53-Ser-15, a downstream
checkpoint protein, was examined. In response to C-1027, ATRdeficient cells phosphorylated both Chk2 and p53 as robustly as
mock siRNA-treated cells (Fig. 5B) (9, 10). Similar to C-1027, no
changes in deschloro-induced phosphorylation of either protein
was observed, regardless of ATR status (Fig. 5B). In contrast,
desmethyl-induced phosphorylation of Chk2 and p53 was nearly
eliminated in ATR-deficient cells (Fig. 5B).
Typically, cells with a diminished capability to activate cell
cycle checkpoints are more susceptible to growth inhibition or
cell death (1). The IC50s for growth inhibition of mock and
siATR-treated cells were determined after treatment with
C-1027, desmethyl, or deschloro. The IC50 of mock-treated cells
was divided by the IC50 of siATR cells to determine a hypersensitivity ratio. Ratios of ⬇1 represent an equal cellular sensitivity in the presence or absence of ATR, whereas a ratio of ⬎1
represents enhanced sensitivity.
As expected from our previous study (10), C-1027 did not
induce significant hypersensitive cell growth inhibition (ratio of
1.1) (Fig. 5C). Similar to C-1027, the growth inhibition ratio of
deschloro was 1.3, consistent with its pattern of cell cycle
checkpoint activation (Fig. 5C). In comparison, the growth
inhibition ratio for desmethyl was significantly higher (⬇2.5),
consistent with its reduced ability to activate DNA damage
responses in the absence of ATR (Fig. 5C). Overall, the PIKK
regulation of DNA damage responses induced by treatment with
C-1027 or its analogs is consistent with whether the DNA lesions
induced are predominately DSBs, ICLs, or a combination of
both.
Although the ability of C-1027 and desmethyl to induce ICLs is
consistent with their induction of ATM-independent DNA
damage responses, the inability of desmethyl to induce DSB
accounts for its dependence on ATR (Fig. 5) (10, 11).
Although ATM and ATR are thought to regulate DNA
damage responses to DSBs or ICLs respectively, there also is a
functional overlap between these kinases (1). For example, after
treatment with IR or calicheamicin, ATR can minimally activate
p53-Ser-15 and Chk2-Thr-68 in the absence of ATM and fully
activate DNA damage responses when the damage is extensive
or if given more time (2, 7, 8, 19). Perhaps limited ICLs or other
DNA lesions that lead to replication stalling are contributing
factors to such ATM-independent DNA damage responses.
Furthermore, ATR’s late regulation of DNA damage responses
to treatments inducing primarily DSBs (8, 19) would be consistent with the typically slower induction kinetics for ICLs than
DSBs (15, 20). Finally, similar to calicheamicin at elevated levels,
we also have observed that, at high drug concentrations, deschloro, deshydroxy, and neocarzinostatin activate ATMindependent damage responses (T.A.B., unpublished data). This
activation is consistent with their relatively limited ability under
cell-free anaerobic conditions to induce limited ICLs (Fig. 2).
However, whether the induction of limited amounts of ICLs by
radiomimetics would factor into ATR regulation of DNA damage responses is difficult to resolve because induction of extensive DSBs could mask their detection.
Predicting whether a radiomimetic has the potential to induce
ICLs is not readily apparent. Although all enediynes induce
cell-free DSBs under aerobic conditions, for ICLs to be generated under anaerobic conditions, the deoxyribose radicals generated upon hydrogen abstraction must react back with the drug
(12). The efficiency of ICL induction, which is thought to depend
on the proximity and steric availability of the enediyne chromophore to the deoxyribose radicals, differs greatly among
enediynes, yet the variation in DNA break to cross-link activity
between C-1027 and its analogs is unexpected given their
minimal structural differences (Fig. 1) (12). However, the efficiency of ICL production also could be influenced by the relative
reactivity of the enediyne chromophore toward the deoxyribose
radicals. Given the high reactivity of phenols toward radicals, the
phenolic OH at C22⬘ could contribute significantly to C-1027’s
ability to couple with the deoxyribose radicals to undergo ICL
formation. Removal of the C22⬘ hydroxyl group, as exemplified
by the deshydroxy analog, may significantly reduce the reactivity
of the chromophore with the radicals. It is unclear why deschloro,
which retains a phenolic group, shows diminished ability to
induce ICLs, but perhaps the removal of chlorine hinders the
ability of the chromophore to interact with the radicals. Finally,
the desmethyl analog contains two phenolic OH functionalities
(i.e., the ␤-amino acid and the benzoxazolinate moieties) (Fig.
1), both of which could potentially facilitate the coupling with
the deoxyribose radicals, consequently generating ICLs with
high efficiency.
The C-1027 concentrations required to initially detect DSBs and
ICLs are comparable (⬇10 nM), which is in agreement with
findings by Goldberg and coworkers (12) that the generation of
DSBs under aerobic conditions are replaced by ICLs when oxygen
is depleted (Fig. 2). However, desmethyl is ⬇50-fold less active as
a DNA strand scission agent (600 nM) than as a cross-linker (Table
1) (11). The fact that desmethyl is equipotent with C-1027 with
regard to ICLs, although comparable DSB induction occurs at
much higher concentrations, might imply that the benzoxazolinate
substitution renders the drug less optimal for oxygen-generated
DNA breaks, but not anaerobic induction of ICLs. In contrast, the
␤-amino acid substitutions, while reducing DSB activity, suppresses
ICL induction (Table 1) (11).
The relative reduction of strand scission activity, compared
with ICL activity, could therefore be correlated with the in17636 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0708274104
creased reactivity of the enediyne chromophore toward the
nascent deoxyribose radicals. Thus, in cells, enediynes with
multiple phenolic OH moieties, such as the desmethyl analog,
could trap all of the nascent deoxyribose radicals for ICL
formation, thereby limiting DSB induction. C-1027 with one
phenolic OH moiety would be in kinetic competition between
ICL and DSB, whereas for deschloro, the loss of the chloro group
may reduce the functionality of the phenolic group, shifting the
DNA damage balance toward DSBs. Compounds with no phenolic OH moiety, such as the deshydroxy analog, would have
little chance to couple with the nascent deoxyribose radicals,
thereby leading to essentially only DSBs.
We discovered that both DSBs and ICLs were induced after
treating SV40 DNA with C-1027, which suggests that intracellular oxygen levels, at least in proximity to the drug-DNA
binding sites, are sufficiently low that deoxyribose radical binding to the enediyne chromophore is competitive with oxygen
radical quenching (5, 12). However, significant DNA breaks
occur at C-1027 concentrations of ⬍10 nM and cross-links are
barely observed at 40 nM, suggesting that the cellular environment favors C-1027-induced DSBs over ICLs (Fig. 3).
Desmethyl induces primarily ICLs in cells, which is in agreement with anaerobic but not aerobic cell-free DNA experiments
(Figs. 2 and 3). It is striking that intracellular DSB activity at
10,000 nM is so limited it is barely detectable, yet ICL activity at
200 nM is on par with C-1027. In contrast, ICL induction by
deschloro was not detected even at high levels, yet DSB activity
decreases compared with C-1027, consistent with the aerobic
cell-free experiments. Thus, it may be possible to rationally
engineer the C-1027 chromophore to produce new generations
of antitumor compounds with fine-tuned abilities to produce
cellular DSBs, ICLs, or a combination of both.
Although the SV40 analysis provides a facile means to simultaneously assess DSBs and ICLs on a common intracellular DNA
target, the lesion frequency on the SV40 target is limited by its
small 5,243-bp genome. Although comet analysis is an indirect
measure of ICLs based on the suppression of IR-induced comet
signals, it provided evidence that cross-linking is occurring at
drug levels consistent with their cytotoxicities. C-1027 induces
concurrent DSBs and ICLs at 0.3 nM, whereas desmethyl ICLs
also are easily detected at 10 nM (Fig. 4 C and D). The 30-fold
difference between the concentrations of C-1027 and desmethyl
where ICLs were first observed approximates the 60-fold difference in growth inhibition, although the DSBs induced by
C-1027 likely also contribute to its cytotoxicity (11). Although it
is clear that the proportion of ICL induction to DSBs dictates
whether ATM, ATR, or both regulate the damage response,
more study is needed to access their relative contributions to cell
growth inhibition.
Using biosynthetically created analogs of the C-1027 radiomimetic enediyne, our study reveals that regulation of cellular
responses to DNA damage can shift from ATM to ATR or use
both kinases in accordance with the proportion of DSBs to ICLs.
Possibly the overlap of ATM and ATR function in responding to
DNA damage may, at least partially, relate to the proportions of
DSBs and ICLs. Remarkably, minor modifications to the
enediyne chromophore dramatically altered the type of DNA
damage, suggesting a rational approach for the design of new
generations of antitumor agents that require the action of a
particular PIKK. The findings about the C-1027 family of
enediynes may have additional relevance for cancer chemotherapy. For example, the hypoxic nature of tumors renders them
resistant to treatment by IR or radiomimetics because DSB
formation at low oxygen levels is suppressed because of the
inhibition of radical quenching by glutathione and other hydrogen donors (21). Currently, we are investigating whether C-1027
and desmethyl ICL activity is enhanced in hypoxic tumor cells,
leading to increased cytotoxicity.
Kennedy et al.
Cell-Free ICL Detection. BamH1-linearized pBR322 plasmid DNA
was incubated with drug in deoxygenated water in a helium
environment within a glove bag for 4 h at room temperature.
Samples were alkaline-denatured, electrophoresed on a 0.8%
agarose gel, stained with ethidium bromide, photographed by
using a Gel Doc XR, and analyzed by Imagequant software
(Molecular Dynamics, Piscataway, NJ).
Cells. HCT116 human colon carcinoma and BSC-1 African green
monkey kidney cells were cultured as described (9, 13).
SV40 Viral Infection and DNA Lesion Detection. BSC-1 cells were
seeded at 2.5 ⫻ 105 cells per ml, infected with SV40 virus for 40 h,
and then drug-treated for 4 h. DNA was isolated and purified as
described previously (13). DNA was electrophoresed on a 0.8%
agarose gel to detect DSBs and ICLs (as described earlier).
Comet Analysis. After a 4-h drug treatment with or without
post-IR treatment, HCT116 cells were analyzed for DNA strand
breaks as previously described. Essentially, cells embedded in
agarose on slides were electrophoresed at 27 V for 25 min at 4°C
under alkaline-denaturing conditions (pH 13), washed in 0.4 M
Tris, and immersed in 100% methanol and then ethanol. Because
smaller DNA fragments migrated through agarose more quickly,
the length of the DNA tail extruding from the nucleus was
proportional to the level of DNA breaks. After ethidium bromide staining, typically 50 cells were scored as follows: 0, intact
comet heads; 1, cells with a slight DNA migration; 2, full-length
comet tails; 3, full-length tail wider than the nucleus; 4, tail
separated from the nucleus (24).
1. Abraham RT (2001) Genes Dev 15:2177–2196.
2. Helt CE, Cliby WA, Keng PC, Bambara RA, O’Reilly MA (2005) J Biol Chem
280:1186–1192.
3. Shiloh Y (2006) Trends Biochem Sci 31:402–410.
4. Goldberg IH, Kappen LS, Povirk LF, Chin DH (1986) Drugs Under Exper Clin
Res 12:495–505.
5. Povirk LF (1996) Mutat Res 355:71–89.
6. Xi Z, Goldberg IH (1999) in Comprehensive Natural Products Chemistry, eds
Barton DHR, Nakanishi K, Meth-Cohn O (Elsevier Science, Oxford), pp
553–592.
7. Ismail IH, Nystrom S, Nygren J, Hammarsten O (2005) J Biol Chem 280:4649–
4655.
8. Uziel T, Lerenthal Y, Moyal L, Andegeko Y, Mittelman L, Shiloh Y (2003)
EMBO J 22:5612–5621.
9. Dziegielewski J, Beerman TA (2002) J Biol Chem 277:20549–20554.
10. Kennedy DR, Beerman TA (2006) Biochemistry 45:3747–3754.
11. Kennedy DR, Gawron LS, Ju J, Liu W, Shen B, Beerman TA (2007) Cancer
Res 67:773–781.
12. Xu YJ, Xi Z, Zhen YS, Goldberg IH (1997) Biochemistry 36:14975–14984.
13. Grimwade JE, Beerman TA (1986) Mol Pharmacol 30:358–363.
Kennedy et al.
siRNA. Oligofectamine was incubated at 30°C for 20 min in the
presence or absence of siRNA targeted against ATR (CCTCCGTGATGTTGCTTGA; Dharmacon, Lafayette, CO) before
addition to HCT116 cells for 48 h as previously described (10).
In previous studies with the same HCT116 cell line, the ATRtargeted sequence was specific for and selective against ATR
(⬎85% decrease in HCT116 cells) (10).
Immunoblotting. After siRNA treatment, HCT116 cells were
incubated with drugs at 37°C for 1 h, and cellular extracts were
prepared for Western analysis as previously described (10).
Essentially, after cell lysis, extracts were cleared by centrifugation, and equal amounts of protein were electrophoresed on
SDS/PAGE and transferred to a PVDF membrane. The membranes were probed with primary antibodies (anti-phosphoChk1-Ser-345, anti-phospho-Chk2-Thr-68, anti-phospho-p53Ser-15, anti-ATR, and anti-␤-actin), followed by secondary
antibodies conjugated with horseradish peroxidase. Protein
bands were visualized by enhanced chemiluminescence and
quantitated by using a Personal Densitometer SI (Amersham
Biosciences, Piscataway, NJ) and Imagequant software.
Growth Inhibition Assay. After siRNA treatment, HCT116 cells
were drug-treated. After a 3-day incubation, cells were counted
on a Coulter counter. Cell growth inhibition was based on a
comparison of the number of treated to nontreated control cells.
We thank Dr. Irving Goldberg for his suggestion to pursue whether
C-1027’s ATM-independent DNA damage response could relate to an
ability to make additional types of celluar DNA lesions, such as the DNA
interstrand cross-links characterized earlier by his laboratory. We also
thank Dr. Y. Li for providing the wild-type S. globisporus strain, Dr. Mary
McHugh for critical reading the manuscript, and Loretta Gawron for
technical assistance. This work was supported, in part, by National
Cancer Institute Grants CA106312 and CA16056 (to T.A.B.), National
Institutes of Health Grants CA078747 and CA113297 (to B.S.), and
National Institutes of Health Training Grant CA09072–30 (to D.R.K.).
14. McHugh MM, Woynarowski JM, Gawron LS, Otani T, Beerman TA (1995)
Biochemistry 34:1805–1814.
15. Woynarowski JM, McHugh MM, Gawron LS, Beerman TA (1995) Biochemistry 34:13042–13050.
16. Almeida GM, Duarte TL, Steward WP, Jones GD (2006) DNA Repair (Amst)
5:219–225.
17. Shiloh Y (2003) Nat Rev Cancer 3:155–168.
18. Pogozelski WK, Tullius TD (1998) Chem Rev 98:1089–1108.
19. Tibbetts RS, Brumbaugh KM, Williams JM, Sarkaria JN, Cliby WA, Shieh SY,
Taya Y, Prives C, Abraham RT (1999) Genes Dev 13:152–157.
20. Scharer OD (2005) ChemBioChem 6:27–32.
21. Ward JF (1990) Int J Radiat Biol 57:1141–1150.
22. Liu W, Christenson SD, Standage S, Shen B (2002) Science 297:1170 –
1173.
23. Van Lanen SG, Dorrestein PC, Christenson SD, Liu W, Ju J, Kelleher NL, Shen
B (2005) J Am Chem Soc 127:11594–11595.
24. Tice RR, Agurell E, Anderson D, Burlinson B, Hartmann A, Kobayashi H,
Miyamae Y, Rojas E, Ryu JC, Sasaki YF (2000) Environ Mol Mutagen
35:206–221.
PNAS 兩 November 6, 2007 兩 vol. 104 兩 no. 45 兩 17637
BIOCHEMISTRY
Materials and Methods
Chemicals. Fermentation, production, isolation, purification, and
identification of C-1027, desmethyl, deschloro, and deshydroxy
from Streptomyces globisporus wild-type and engineered strains
were carried out as described (22, 23).