Clofarabine 5-di and -triphosphates inhibit human
ribonucleotide reductase by altering the quaternary
structure of its large subunit
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Aye, Y., and J. Stubbe. “Clofarabine 5’-di and -triphosphates
inhibit human ribonucleotide reductase by altering the quaternary
structure of its large subunit.” Proceedings of the National
Academy of Sciences 108.24 (2011): 9815-9820. ©2011 by the
National Academy of Sciences.
As Published
http://dx.doi.org/10.1073/pnas.1013274108
Publisher
National Academy of Sciences (U.S.)
Version
Final published version
Accessed
Wed May 25 21:49:09 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/67468
Terms of Use
Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
Clofarabine 5′-di and -triphosphates inhibit
human ribonucleotide reductase by altering
the quaternary structure of its large subunit
Yimon Ayea and JoAnne Stubbea,b,1
Departments of aChemistry and bBiology, Massachusetts Institute of Technology, Cambridge, MA 02139
Edited by Barry S. Cooperman, University of Pennsylvania, Philadelphia, PA, and accepted by the Editorial Board May 3, 2011 (received for review September
4, 2010)
allosteric regulation ∣ oligomerization ∣ slow-binding inhibitor
C
lofarabine (ClF, 2-chloro-2′-arabino-fluoro-2′-deoxyadenosine, Clolar®, 1) is a nucleoside that is used to treat pediatric
acute leukemias (1–3). Its mode of cytotoxicity is proposed to be
similar, with several important distinctions, to gemcitabine
(F2 C, 2), a drug used in the treatment of advanced pancreatic
cancer and nonsmall cell lung carcinomas (4).
NH2
N
HO
H
HO
O
NH2
N
N
N N
F
Cl
HO
1
H
H
HO
O
N O
F
2
F
Both nucleosides enter cells via the CNT- or ENT-type transporters and their phosphorylation is required for cytotoxicity
(1–4). The monophosphate of ClF (ClFMP) is generated predominantly by deoxycytidine kinase and sequentially converted to
the diphosphate (ClFDP) (the rate limiting step) (5–7) and the
triphosphate (ClFTP) by MP and DP kinases, respectively (6–8).
A key step in the metabolism of both ClF and F2C is proposed to
be inhibition of ribonucleotide reductase (RNR) (4, 7–9), the
enzyme responsible for conversion of NDPs to deoxynucleoside
diphosphates (dNDPs) in all organisms. F2 CDP acts as a substoichiometric mechanism-based inhibitor (MBI) to inactivate
human RNR (hRNR) (4, 10), whereas ClFTP is proposed to inactivate hRNR by binding to an allosteric site that binds dATP
www.pnas.org/cgi/doi/10.1073/pnas.1013274108
(9). In both cases, depletion of dNDP pools results in reduction
of dNTP pools, allowing ClFTP or F2 CTP to compete effectively
with the available dNTPs for incorporation into DNA. For ClFTP,
the consequences are dependent on the ClFTP/dATP ratio and
the inhibition of DNA polymerase α (7, 8). Chain termination of
the polymerization process triggers apoptosis (5). Our recent
results established that both fluoride ions (F− ) of F2 CDP are
eliminated during the covalent modification of RNR (4, 10,
11). We thus proposed that ClFDP might function as an MBI with
F− loss, furanone formation, and covalent enzyme modification
(10–14) and hence both the di- and triphosphates might play an
important role in RNR inhibition. This paper reports characterization of ClFDP and -TP (Scheme S1) with the hRNR involved
in DNA replication and describes an unprecedented mode of inhibition of this enzyme.
RNRs supply monomeric precursors required for DNA replication/repair and are pivotal in controlling the fidelity of these
processes (15). hRNRs are composed of α- and β-subunits that
form an active quaternary complex(es) whose composition(s) is
(are) currently under investigation (16–19). There are two different β-subunits: One is involved in DNA replication (the focus of
this paper); the other, p53β, is induced by p53 and is proposed to
be involved in supplying dNTPs for mitochondrial DNA replication and repair (20). α binds four substrates (UDP, CDP, ADP,
and GDP, C site) and has two allosteric effector binding sites,
the specificity site (S site) and the activity site (A site). The former controls which substrate is reduced by binding the appropriate (d)NTP (ATP, dATP, dGTP, and TTP). The latter, also called
the ATP cone domain, controls NDP reduction rate: When ATP
is bound, RNR is active and when dATP is bound, RNR is inactive. The β-subunit houses the essential diferric-tyrosyl radical
(Y•) cofactor that initiates nucleotide reduction in the active site
of α by a 35-Å proton-coupled electron transfer process (21). Interactions between α and β are weak and modulated by NDP and
dNTP/ATP binding to α with the active forms of the mammalian
RNR proposed to be αn βm (n, m ¼ 2 or 6) (16–19) and the inactive form produced by dATP binding to the A site, proposed
to be α6 β2 (17, 19).
Studies on hRNR in vitro are challenging. First, Y• in β has
t1∕2 ¼ 25 min at 37 °C. All inhibition studies must be corrected
for this inherent instability. In addition, the weak interaction
of the two subunits (18) and the multiple quaternary structures
of both α and the αβ-complex (4, 16–19) make the concentrations
at which the inhibition studies are carried out important.
Author contributions: Y.A. and J.S. designed research; Y.A. performed research; Y.A.
analyzed data; and Y.A. and J.S. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission. B.S.C. is a guest editor invited by the
Editorial Board.
1
To whom correspondence should be addressed. E-mail: stubbe@mit.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/
doi:10.1073/pnas.1013274108/-/DCSupplemental.
PNAS ∣ June 14, 2011 ∣ vol. 108 ∣ no. 24 ∣ 9815–9820
BIOCHEMISTRY
Human ribonucleotide reductases (hRNRs) catalyze the conversion
of nucleotides to deoxynucleotides and are composed of α- and
β-subunits that form active αn βm (n, m ¼ 2 or 6) complexes. α binds
NDP substrates (CDP, UDP, ADP, and GDP, C site) as well as ATP and
dNTPs (dATP, dGTP, TTP) allosteric effectors that control enzyme
activity (A site) and substrate specificity (S site). Clofarabine
(ClF), an adenosine analog, is used in the treatment of refractory
leukemias. Its mode of cytotoxicity is thought to be associated
in part with the triphosphate functioning as an allosteric inhibitor
of hRNR. Studies on the mechanism of inhibition of hRNR by ClF
di- and triphosphates (ClFDP and ClFTP) are presented. ClFTP is a
reversible inhibitor (K i ¼ 40 nM) that rapidly inactivates hRNR.
However, with time, 50% of the activity is recovered. D57N-α, a
mutant with an altered A site, prevents inhibition by ClFTP,
suggesting its A site binding. ClFDP is a slow-binding, reversible
inhibitor (K i ¼ 17 nM; t 1∕2 ¼ 23 min). CDP protects α from its inhibition. The altered off-rate of ClFDP from E • ClFDP by ClFTP
(A site) or dGTP (S site) and its inhibition of D57N-α together implicate its C site binding. Size exclusion chromatography of hRNR or
α alone with ClFDP or ClFTP, ATP or dGTP, reveals in each case that
α forms a kinetically stable hexameric state. This is the first example of hexamerization of α induced by an NDP analog that reversibly binds at the active site.
Results
Specific Activity of Human ðHisÞ6 -α, -β, and Untagged-α. Talon affinity
chromatography replaced Ni-NTA for ðHisÞ6 -α purification (4)
increasing yield and purity (Fig. S1). β was isolated in apo form
and reconstituted in vitro giving 0.8–1.2 Y • ∕β2 and 3.5 Fe∕β2
(4). Typical specific activities (SAs) for each ðHisÞ6 -tagged subunit, with saturating amounts of the second subunit using CDP/
ATP were 650–890 nmol∕ðmin ·mgÞ for α and 3,000–4,100 for
β. For ADP/dGTP/ATP the values were 140–160 for α and
200–220 for β. Without ATP, the values dropped to 70–80 for
α and 80–95 for β. SAs of each tagged subunit for CDP or
ADP reduction with 3 mM ATP remain within these ranges over
5 nM–5 μM of α (β) with 20–5-fold excess β (α). The corresponding SA of the untagged-α, under identical conditions, is increased
approximately 1.5-fold relative to tagged-α.
Time-Dependent Inactivation of hRNR by ClFDP. Our previous studies
with 2′-substituted -2′-deoxynucleotides such as 2′-ribofluoro-2′deoxycytidine 5′-diphosphate and F2 CDP showed that both compounds were MBIs of bacterial and hRNRs (4, 10–14). Studies on
Lactobacillus leichmannii RNR demonstrated that 2′-arabinohalonucleotides also could function as MBIs (12). Our hypothesis
was that ClFDP might function similarly with removal of the 3′hydrogen resulting in F− loss (13, 14). We initially focused on
whether ClFDP could function as a time-dependent inhibitor
of dNDP formation. Because ClF is an adenosine analog, dGTP
and ATP were used as allosteric effectors as they maximize ADP
reduction and potentially ClFDP inhibition. Incubation of one
equivalent of ClFDP with hRNR∕dGTP∕ ATP, followed by assaying for residual activity of α using [5-3 H]-CDP, ATP and saturating β, resulted in time-dependent inhibition (Fig. 1A), with
complete activity loss by 20 min. Many MBIs cause inactivation
9816 ∣
www.pnas.org/cgi/doi/10.1073/pnas.1013274108
B 100
80
80
% Activity
A 100
% Activity
Our studies show that ClFDP is a slow-binding, reversible
inhibitor with a K i of 17 nM and a t1∕2 of 23 min that is not modified by hRNR. They also show that ClFTP is a reversible inhibitor (K i ¼ 40 nM) that rapidly inactivates hRNR, but regains
approximately 50% activity with time. A series of experiments
was undertaken to determine the site(s) of binding of ClFDP
and -TP to α responsible for inhibition. First, studies are carried
out with D57N-α (altered A site), which is no longer inhibited by
dATP (16, 22). They reveal that ClFTP, up to 6 μM (150 × K i ), is
unable to inhibit hRNR, suggesting that it binds only to the A site.
Previous studies in crude cell extracts have shown that ClFTP at
0.3–10 μM completely inhibits CDP reductase activities (5–9).
The same mutation, however, has no effect on ClFDP inhibition,
suggesting that it does not bind to the A site. Second, a series of
preincubation–dilution experiments with ClFTP and either CDP
(C site) or dGTP (S site) shows no effect on its inhibition, whereas
similar experiments with ClFDP and CDP show mutually exclusive binding, implying ClFDP binds to the C site. Third, studies
measuring ClFDP release rate from [8-3 H]-ClFDP•hRNR in the
presence of dGTP or ClFTP show rate constants significantly different from that in their absence, providing additional support for
C-site specificity.
Finally the mode of inhibition of both compounds is associated
with altered quaternary structures of α. Size exclusion chromatography (SEC) subsequent to incubation of the hRNR or α alone
with each inhibitor in the presence or absence of effectors shows
that inhibition is associated with ClFDP- or ClFTP-dependent
and effector-independent protein oligomerization. In both cases,
the quaternary structure changes to α6βn or α6 (n ≤ 2) that is
more kinetically stable than dATP-induced hexamers. These studies identify ClF nucleotides as potent reversible inhibitors of
hRNR and demonstrate the previously unrecognized important
role of ClFDP. Furthermore, they provide evidence that NDP’s
binding to the C site of hRNR large subunit can alter quaternary
structure.
60
40
20
60
40
20
0
0
5
10
15
20
0
0
time (min)
5
10
15
20
time (min)
Fig. 1. (A) Time-dependent inhibition by ClFDP. Inhibition mixtures (IMs)
contained 1.2 μM ½α ¼ ½β ¼ ½ClFDP, 0.1 mM dGTP, and 0 (♦) or 3 ( ) mM
ATP (in assay for α); 0 (▪) or 3 (▴) mM ATP (for β). (▾) was identical to ( )
except it contained 6.0 μM ClFDP. (○) was identical to (▾) except IM contained
D57N-α in place of WT-α. (B) Initial nearly complete inhibition by ClFTP and
subsequent recovery of approximately 50% activity. IM contained 1.2 μM
½α ¼ ½β, 1.2 ( ), 6.0 (▪) or 12 (♦) μM ClFTP (in assay for α), 1.2 (▾, ○) μM ClFTP
(for β), and 0 (▪,♦,▾) or 3 ( , ○) mM ATP. D57N-α CDP reductase activity was
unaffected (▴) under identical conditions to ( ).
•
•
•
•
•
of RNR in part by loss of the Y• (14). Thus, the reaction was also
assayed for β-activity (Fig. 1A). Minimal loss of activity was observed. The studies show that ClFDP is a potent inhibitor of α.
Analysis for Chloroadenine (ClA) Release from ClFDP Incubated with
hRNR. One of the hallmarks of 2′-substituted MBIs is cleavage
of the nucleobase from the sugar and loss of PPi , generating a
furanone species that alkylates the protein (14). To deterimine
if ClA is released, hRNR was incubated with a stoichiometric
amount of ClFDP for 20 min, conditions in which complete inhibition occurred (Fig. 1A). Protein was then removed by ultrafiltration and the filtrate analyzed by HPLC (Fig. S2A). No ClA
was detected. To establish that the inhibitor remained unchanged,
the enzyme was thermally denatured and removed by centrifugation. The supernatant was treated in two ways. First, it was incubated with alkaline phosphatase and then analyzed by HPLC. ClF
was recovered quantitatively (Fig. S2B). Second, supernatant was
lyophilized and then shown to inhibit the enzyme stochiometrically. The results indicate that ClFDP is not a mechanism-based,
but a slow-onset, reversible inhibitor.
ClFDP Is a Slow-Binding, Reversible Inhibitor. The upper limit for koff
for ClFDP from α•ClFDP was first estimated to be approximately 0.1 min−1 by measuring recovery of α-activity upon dilution (see SI Text p. 3 and Fig. S3). Subsequently, progress curve
analyses of α-activity in the presence of 3 mM [8-14 C]-ADP and
variable [ClFDP] were carried out (Fig. 2A). In the absence of
inhibitor, dADP formation is linear over 15 min (with adjustments to account for enzyme instability) at 37 °C (controls A
and B, SI Text). In the presence of ClFDP, kinetics are nonlinear:
with an initial rapid phase (velocity νi ), and ensuing slower steady-state phase (velocity νs ). This biphasic profile is commonly observed with nonclassical slow-onset reversible inhibitors (23–25).
This type of inhibition can be described as a single-step mechanism [Scheme 1(i)], where the rate constant for association (k1 ) or
dissociation (k−1 ) or both are inherently slow. Alternatively it can
be described as a two-step mechanism involving rapid equilibrium
binding of I to E giving an E • I complex, followed by a slow isomerization to a higher affinity complex, E • I [Scheme 1(ii)],
where the rate constants k−2 and k2 are both low, but k−2 ≪
k2 . To distinguish between (i) and (ii), a nonlinear regression analysis of the progress curves in Fig. 2A was performed using Eq. 1
to obtain the best fit values for the observed rate constant, kobs ,
for onset of inhibition for each [I].
Aye and Stubbe
B
0.4
0.2
20
10
0
0.0
3
6
9
[I] ( M)
A
12
B
C
app
release buffer
C
K i ¼
1,000
[dCDP] (nM)
600
400
Characterization of hRNR Inhibition Mechanism by ClFTP: A Rapid
Reversible Inhibitor. Previous studies in cultured K562 and CEM
200
0
0
0
5
10
0
15
time (min)
2
4
6
time (min)
Fig. 2. ClFDP is a C-site-directed, slow-binding reversible inhibitor of hRNR.
(A) Progress curves for ADP reduction. [ClFDP] was varied from (★) 0, (•) 1.5,
(▴) 3.0, (♦) 6.0, (▪) 9.0 to (▾) 12.0 μM. The solid curves are the best nonlinear
fits of the data to Eq. 1. The dotted line is the uninhibited progress curve.
(Inset) A replot of kobs for the progress curves as a function of [ClFDP].
The hyperbolic curve is the best nonlinear fit of the data to Eq. 2. (B) Perturbations of [8-3 H]-ClFDP residence t 1∕2 in response to ternary complex formation. Release buffer contained: 0.1 mM dGTP (A); 0 or 5 μM ClFDP (B); and
5 μM ½ClFTP ¼ ½ClFDP (C). (C) CDP protects α from ClFDP inhibition. The preincubation mixture contained hRNR (5 μM) and either CDP alone [0.5 (▴) or 3
( ) mM], or ClFDP alone [10 μM (▾)], or both [10 μM ClFDP and 0.5 (▪) or
3 (♦) mM CDP].
•
ðνi − νs Þð1 − γÞ 1 − γðe−kobs t Þ
;
ln
½P ¼ νs t þ
kobs γ
ð1 − γÞ
where
γ¼
½E
ν 2
1− s :
νi
½I
kobs ¼ k−2 þ
k2 ½I
app
K i þ ½I
:
k1
k−1
(ii)
E I
E + I
k1
k−1
55
44
B
33
22
100
11
10
80
5
60
40
20
0
0
0.8
1.6 2
4
6
[ClFTP] ( µM)
0
0
E I
k2
k−2
E I
*
Scheme 1. Two possible kinetic mechanisms that describe slow-binding inhibition.
Aye and Stubbe
15
[2]
app
E + I
•
A
The best fit values of k2 and K i are 0.58 0.01 min−1 and
0.40 0.06 μM, respectively. The value for k−2 (0.03
0.01 min−1 ) is obtained from the y intercept (Fig. 2A, Inset)
and is associated with substantial error due to the inability to
(i)
cells suggested that RNR is a target responsible in part for ClF’s
cytotoxicity. Their measurements of the relative concentrations
of ClFDP and ClFTP (approximately 1∶7) (8), altered dNTP
pools (6–9), and inhibition of NDP reductase activity in the crude
cell extracts by ClFTP (7, 8) led to the suggestion that the triphosphate was responsible for RNR inhibition. To test this proposal
in vitro with purified hRNR, time-dependent inhibition studies
were carried out (Fig. 1B). They showed that with a 5-fold excess
of ClFTP over RNR, inactivation of α was complete within the
first time point. The inhibited hRNR•ClFTP complex was then
analyzed using Sephadex-G25 chromatography or ultrafiltration
(Fig. S2C, ▪, ). In both cases, 100% activity was recovered demonstrating that ClFTP is a rapidly reversible inhibitor, an observation in contrast to ClFDP (▴). As with ClFDP, a similar
inhibition experiment probing the remaining β-activity showed
that ClFTP also primarily targets α (Fig. 1B).
Rapid reversible interaction between hRNR and ClFTP was
also apparent from the linearity of the progress curves at [ClFTP]
between 0 to 300 nM and ½E ¼ 300 nM (Fig. 3A). However, with
600 to 1,500 nM ClFTP, a lag phase was observed, with no additional increase in inhibition above 600 nM (Fig. 3A). This unusual
[1]
In Eq. 1, [P] is [dADP], [E] is the total enzyme concentration, and
t is time. Tight-binding inhibition under these conditions is accounted for by the γ-term of Eq. 1, which is required when [I]
is similar to [E] (Fig. 2A) (23–25). A plot of kobs vs [I] gives
the results in Fig. 2A, Inset. The hyperbolic relationship suggests
that ClFDP inhibits hRNR via mechanism (ii), Scheme 1. Subseapp
quent analysis using Eq. 2 (23–25) gives k2 , k−2 , and K i , where
app
K i is an apparent K i for the initial collision complex E • I
(or k−1 ∕k1 ):
[3]
The value of 17 7 nM obtained is within the range required
to observe ClFDP release in the off-rate experiment above. Thus
two independent approaches support the slow-binding inhibitor
model (ii) (Scheme 1) and provide a binding constant of 17 nM.
800
200
Ki
:
1 þ ðk2 ∕k−2 Þ
5
10
15
20
time (min)
Fig. 3. ClFTP rapidly and reversibly inactivates human α, approaching maximum inhibition of approximately 50%. (A) Progress curves for CDP reduction. [ClFTP] was varied from (♦) 0, (•) 150, (▪) 300, (▴) 600, (▾) 900, (♦)
1,200 to (○) 1,500 nM. The solid lines are the best nonlinear fits of the data
to straight lines through the origin. The fitted linear lines were omitted for
½ClFTP ≥ 600 nM for clarity. (Inset) A plot of fractional inactivation as a function of [ClFTP]. Each data point was derived from the respective slope generated by nonlinear fitting of the data in A to straight lines. The dotted line
represents the best nonlinear fit to Eq. 4. (B) Activity of WT-α ( ) vs. D57N-α
(♦) as a function of [ClFTP]. dCDP production rates with 0–6.0 μM [ClFTP],
0.3 μM α, 3.0 μM β, 2 mM [5-3 H]-CDP, and 3 mM ATP measured over 20 min.
•
PNAS ∣
June 14, 2011 ∣
vol. 108 ∣
no. 24 ∣
9817
BIOCHEMISTRY
0
% Activity
400
umol dCDP produced
per mg of H1
kobs (min-1)
[dADP produced] ( M)
[8-3H]-ClFDP t1/2 (min)
0.6
A 600
make measurements at low [I] (ClFDP). This value, however, lies
within the range of the upper limit for koff of approximately
0.1 min−1 measured above. From k−2 , the α•ClFDP complex
t1∕2 ¼ 23 3.8 min (23–25).
Assuming that the rate limiting step in (ii) (Scheme 1) is associated with the breakdown of E • I to E • I, the true affinity of
ClFDP or overall dissociation constant of E • I (K i ) can be
calculated using Eq. 3 (23–25):
30
behavior was more apparent in a replot of fractional activity
of RNR vs. [I] (Fig. 3A, Inset), where maximum inhibition is
approximately 50%. This interesting behavior will be discussed
subsequently. A K i of 40 nM was extracted from the data in
Fig. 3A, Inset using Eq. 4, which imposes no restrictions about
[I] relative to [E] (23–26).
Y¼
50
½ð½E þ ½I þ K i Þ − fð½E þ ½I þ K i Þ2 − 4½E½Ig1∕2 : [4]
½E
In this equation, Y is the percent of fractional inhibition, [E]
and [I] are the total concentrations of α and ClFTP, respectively,
and K i is the dissociation constant.
Isolation/Characterization of ðHisÞ6 -D57N-α: A Probe for ClFTP Site
Specificity. RNR has two well-characterized effector binding sites:
the A site governing activity and the S site controlling specificity
(15). ClFTP can potentially bind to either or both sites, if it is a
dATP analog as initially proposed (9). dATP binding is associated
with changes in the quaternary structure of mouse RNR to an
aggregate proposed to be α6 or α6β2 in the absence or presence
of β2, respectively (17, 18). D57 resides in the A site (22) and the
mutation of D57 to N, in the mouse α (D57N-M1) substantially
decreases the dATP binding affinity to this mutant, essentially removing its ability to inhibit RNR (16, 22). We hypothesized that
the human D57N mutant would have the same phenotype as
mouse and that ClFTP, under conditions in which the WT α is
100% inhibited, would no longer inhibit α.
To investigate this proposal, the human D57N-α was overexpressed in Escherichia coli in soluble form only with growth
at 18 °C. The mutant was purified to >95% homogeneity
(Table S1 and Fig. S1) with a typical yield of 0.3 mg∕g of cell
paste. The mutant was no longer inhibited by increasing concentrations of ½dATP ATP (compare Fig. S4 A and B), an outcome previously reported for the mouse counterpart (16, 22).
Furthermore, ADP reductase activity in the presence of dGTP,
an S-site effector, was 45% of WT-α, whereas CDP reductase activity in the presence of ATP was 65% WT-α. The former result
demonstrates that the S site on the mutant remains functional,
whereas the latter demonstrates the enzyme is minimally affected
by the mutation. Finally, the mutant enzyme is not inhibited by
ClFTP under conditions where the WT is completely inhibited
(Figs. 1B and 3B). These data indicate inhibition arises from preferential binding of ClFTP to the A site because increasing the
[ClFTP] to 6 μM (150 × K i ) showed no additional inhibition of
mutant-α (Fig. 3B).
Time-Dependent Inactivation Assays with ClFTP Reveal Transient Inhibition of α Followed by Approximately 50% Recovery of Activity.
Time-dependent inactivation with 5-fold molar excess ClFTP:α,
analyzed over 20 min [conditions where D57N-α remains uninhibited (Fig. 3B)], revealed rapid loss of >90% activity followed
by recovery of activity to approximately half the maximum
(Fig. 1B). These results may provide an explanation for the lag
phase associated with the initial phases of the progress curves
(Fig. 3A). Our hypothesis is that the lag phase observed in the
first 5 min corresponds to the transient, >90% inactivated α
(Fig. 1B) associated with binding of one ClFTP/α and that with
time, the enzyme changes conformation such that the inhibitor no
longer binds to all available A sites, giving α with approximately
50% activity (Fig. 1B). In principle, ClFTP could bind to both S
and A sites (2 ClFTP/α) with subsequent loss from the S site.
However, the observation that no initial inhibition occurs upon
incubation of D57N-α with ClFTP (Fig. 1B, ▴) under identical
conditions to WT-α (Fig. 1B, ) suggests that if ClFTP does bind
to the S site, it is not inhibitory. Titration experiments monitoring
dCDP production under initial and final conditions (2 vs. 20-min
incubation times, respectively) provide compelling evidence for
two different inhibited states of α by ClFTP (Fig. S4C). Binding
•
9818 ∣
www.pnas.org/cgi/doi/10.1073/pnas.1013274108
studies with isotopically labeled ClFTP are required for detailed
understanding of stoichiometry associated with these unusual
results.
Inhibition of ðHisÞ6 -D57N-α by ClFDP. The diphosphate is expected to
bind to the C site. Nevertheless, it has been reported that natural
dNDPs at low mM concentrations can act as allosteric effectors
(27). In addition, the presence of F and Cl could enhance affinity
for the A site, despite the weakened binding anticipated by replacing the triphosphate (dATP and ATP normal effectors) with the
diphosphate. The results shown in Fig. 1A (○) reveal that unlike
ClFTP, ClFDP inactivates D57N-α similarly to WT-α. The data
thus suggest that ClFDP does not bind to the A site.
Perturbation of k off of [8-3 H]-ClFDP from E • I (Scheme 1) by dGTP and
ClFTP. To further confirm ClFDP binding to the C site, hRNR and
[8-3 H]-ClFDP (½E ¼ ½I ¼ 2 μM) were preincubated to form E •
I complex (Scheme 1), which was then rapidly diluted
400- or 40-fold into the release buffer (RB) containing 0 or
5 μM (approximately 300 × K i ) ClFDP, respectively, and the
koff (s) of [8-3 H]-ClFDP measured. The presence of ClFDP in
RB minimizes [8-3 H]-ClFDP reassociation with enzyme. The
measurements were then repeated with either saturating dGTP
[0.1 mM, approximately 50 × K i (16, 28)], or ClFTP (5 μM,
125 × K i ) in addition to ClFDP, in the RB. Any alterations in
measured koff s within each pair are most reasonably attributed
to ternary complex formation wherein ClFTP or dGTP binds simultaneously to [8-3 H]-ClFDP•E [E • I , Scheme 1 (ii)].
The koff s for [8-3 H]-ClFDP were first measured ClFDP in
the RB and found to be similar: 0.026 vs. 0.034 min−1 with
t1∕2 , 26 vs. 21 min [Fig. 2B, RB ¼ B, and Fig. S5 A (▴) and B
(▾)], consistent with ClFDP binding exclusively to the C site, because binding of ClFDP to other sites (S or A site) might have
altered the [8-3 H]-ClFDP release rate. The koff value is also within range of that extracted from progress curve analysis of ClFDP
(Fig. 2A). In contrast to these results, when a saturating amount
of dGTP (S site), or ClFTP (A site) in addition to ClFDP, was
placed in the RB, increases in koff of 6- and 3.5-fold, respectively,
were observed [Fig. 2B, RB ¼ A and C vs. B, and Fig. S5, ( ) vs.
(▾), and (▪) vs. (▴)]. Thus ClFDP binds at a site unique from
these two nucleotides, supporting C-site binding.
•
CDP Protection Against ClFDP Inhibition of α: Preincubation-Dilution
Experiments. In a third set of experiments to support ClFDP bind-
ing to the C site, a series of pairwise preincubation–dilution assays was undertaken. These experiments were modeled after
those described by Nakamura and Abeles to determine the site
specificity of slow-onset and slow-release reversible inhibitors
such as compactin for HMGCoA reductase, when high substrate
concentrations render traditional substrate protection assays ineffective due to substrate inhibition (29). For hRNR the SA of α
at 0.5 mM CDP, 15 mM Mg2þ , and 3 mM ATP is reduced by half
when [CDP] is increased to 10 mM, and increasing [Mg2þ ] does
not prevent the observed substrate inhibition.
We initially tested the preincubation–dilution assays with
ClFTP against CDP and dGTP. As expected for a rapid reversible
inhibitor, dilution of the preincubation mixture and analysis for
dCDP formation showed that neither CDP (C site) nor dGTP
(A site) had any effect on inhibition by ClFTP (Fig. S6 A and B).
We then studied the effect of CDP on the slow-onset inhibition
of ClFDP (Fig. 2C). α (5 μM) was preincubated with ClFDP
(10 μM ∼ 600 × K i ) alone, with CDP (0.5 or 3.0 mM) alone, or
with ClFDP (10 μM) and CDP (0.5 or 3.0 mM). CDP thus protects α from ClFDP, consistent with the expected competition
between CDP and ClFDP for binding to the C site.
Quaternary Structure of hRNR Subsequent to Inactivation by ClFDP,
ClFTP or ClFDP and ClFTP. Previous studies have shown that dATP
Aye and Stubbe
Table 1. ClFDP and ClFTP-induced hRNR oligomerization analyzed
by SEC
Entry
Protein
(effector/inhibitor)*,†
Retention
time, min‡
Apparent
mass,
kDa§,¶
For entries 1–14, elution buffer contained 0.5 mM ATP:
1
α, β (dGTP, ClFDP)
17.4
580
2
α, β (dGTP, ClFTP)
17.1
621
3
α, β (ClFTP)
17.1
621
4
α, β (dGTP, ClFDP, ClFTP)
17.2
608
5
α, β (dGTP)
25.4
—
6
α, β (dGTP, ATP)
24.5
—
7
α (dGTP, ClFDP)
17.6
553
8
α (dGTP)
22.4
181
9
α (ClFTP)
17.5
567
10
α
24.0
125
11
α (dATP)
20.7
134
12
untagged-α (dGTP, ClFDP)
17.7
541
13
untagged-α (ClFTP)
17.6
553
14
untagged-α
24.2
—
For entries 15–19, elution buffer contained no effectors:
15
α (dGTP, ClFDP)
17.7
541
16
α (ClFDP)
17.7
541
17
α (ClFTP)
17.5
567
18
α
26.0
78
19
α (dATP)
23.3
—
For entry 20, elution buffer contained 20 μM dATP:
20
α (dATP)
17.9
516
Fig. S7∥
A
B
—**
C
—††
F
D
G
E
I
H
O
P
Q
J
K
L
M
—‡‡
N
*Elution buffer: 150 mM NaCl, 15 mM MgCl2 in 50 mM Hepes (pH 7.6) ±
specified effectors. Flow rate ¼ 0.5 mL∕ min.
†
Incubation sample contained ½ðHisÞ6 -α ¼ ½ðHisÞ6 -β ¼ 15 mM; ½untagged-α ¼
15 μM; ½ClFDP ¼ ½ClFTP ¼ 75 μM; ½dGTP ¼ 0.1 mM; ½ATP ¼ 3 mM; and
½dATP ¼ 0.6 mM (entry 11) or 20 μM (entry 19). Parentheses specify
nucleotides in each sample.
‡
Of major peak. Differences (if any) between duplicate runs are less than
0.05 min. In cases with ClFTP, elution profiles and RTs are identical
between 2 and 20 min incubation.
§
Determined by comparison with GE MW standards (Fig. S7, Inset); provided
only for single, well-defined sharp peaks (see Fig. S7 A–Q).
¶
Theoretical mass: αn (n, kDa): (1, 92); (2, 184); (4, 368); (6, 553); (7, 645). α6βm
(m, kDa): (1, 600); (2, 647); (4, 740); (6, 834). Untagged-α (n, kDa): (1, 90);
(2, 180); (4, 360); (6, 540); (7, 630).
∥
Respective elution profiles in Fig. S7 A–Q.
**Profile of entry 3 is identical to that of entry 2 (Fig. S7B).
††
Profile of entry 5 looks identical to that of entry 6 (Fig. S7F).
‡‡
Profile of entry 19 looks identical to that of entry 11 (Fig. S7H) except RT at
maximum absorbance is shifted.
Aye and Stubbe
Finally, a sample containing α with 0.6 mM dATP, where both
the S and A sites are saturated (16, 28) and approximately 90%
enzyme activity is lost (Fig. S4A), also elutes with broad features
suggesting an equilibrium mixture of α oligomers (entries 11
and 19, Fig. S7H). Only when dATP (20 μM) is included in the
elution buffer is the recently reported dATP-induced hexamerization of hRNR α detected (Fig. S7N) (19).
The results with ClFDP and ClFTP under various conditions
with α alone or holoenzyme are in stark contrast. In all cases, the
protein(s) elute(s) as a sharp peak and the RTs in general vary
from 17.1 to 17.7 min (Table 1). The RTs are nearly independent
of the presence of dGTP or ATP in the sample or the elution buffer. They are also nearly the same with α or holoenzyme in the
sample, although they are slightly longer in the former case.
Molecular weight (MW) standards (Fig. S7, Inset) give apparent
MWs from 540 to 620 kDa within the range of α6 to α6βn where
n ¼ 0–2 (Table 1), respectively. SDS-PAGE analysis of proteincontaining fractions eluted from the column (corresponding to
entries 1–4, Table 1) relative to standards having different ratios
of α and β, demonstrated that the 17-min peak constitutes an α6
complex with variable amounts of βn (n ≤ 2). Thus neither β nor
any effector (at S or A site) is required for hexamerization.
Two points are of interest. First, dGTP binding to the S site, which
enhances α dimerization (16), is not a prerequisite for hexamerization caused by ClFDP (Fig. S7 J vs. K). Second, unlike dATPinduced hexamerization, which is observed only by SEC with
20 μM dATP in the elution buffer (entries 11 and 19 vs. 20)
(16, 19), ClFTP, which binds to the same site, yields a unique
form of α6 that is much more kinetically stable (Fig. S7 E and
L vs. H and N).
Finally, because α-hexamerization is thought to involve its
N terminus (the ATP cone domain), the ðHisÞ6 tag and linker
were cleaved with thrombin to generate “untagged”-α with an
additional Gly-Ser unit. A number of experiments were repeated
(entries 12–14, Table 1) with untagged-α, and the RTs remain
unaltered after taking into account the change in MW.
The ClFDP SEC studies describe a reversible NDP inhibitor
binding at the active site causing α-hexamerization. The studies
collectively suggest that inhibition by ClFDP or ClFTP is related
to conversion of α to α6, which can then bind β subunit(s). The
results are intriguing as during the SEC analysis, neither inhibitor
is present in the elution buffer and ClFTP has likely dissociated
and ClFDP partially dissociated from RNR subsequent to sample
injection. Inhibition by ClF nucleotides thus results in different
types of hexameric states of the hRNR large subunit that may
preclude β from binding or binding in an active form.
Discussion
Active and inactive quaternary structures of mouse and Saccharomyces cerevisiae RNR have recently received much attention
(16–19). The αn -subunit has been shown to exist in multiple
oligomeric states (n ¼ 1, 2, and 6). Similarly, the RNR complex,
αn βm , can also exist in multiple states. As thoughtfully described
by Kashlan and Cooperman and subsequently by Rofougaran
et al., the distribution of states will depend minimally on the
concentration of the subunits, their localization, and the concentrations of ATP, dNTPs, and Mg2þ (16, 17). Thus reported differences in quaternary structure are in part related to the conditions
under which the measurements are made and their resolution.
The importance of the quaternary structure of α in the inhibition of RNR in the presence of dATP was initially described for
the E. coli enzyme (15) and has since been found to be universal.
Studies using gas-phase electrophoretic-mobility macromolecule
analysis, electrospray ionization mass spectrometry, analytical
ultracentrifugation (AUC), SEC (17, 18), and recent structural
analyses (19) have all suggested mouse α can generate a hexameric, inactive structure.
PNAS ∣
June 14, 2011 ∣
vol. 108 ∣
no. 24 ∣
9819
BIOCHEMISTRY
inactivates eukaryotic class Ia RNRs by altering the quaternary
structures of α and the holocomplex: in the former case to an
α6 state and in the latter case to an α6β2 state (17, 19). Given
that ClFTP may be an analog of dATP, it is conceivable that
its mechanism of RNR inhibition is associated with alteration
of its quaternary structure. In addition, recent studies with
F2 CDP indicate that this analog not only modifies α covalently
but also alters its quaternary structure (4).
SEC with a General Electric (GE) Superdex™ 200 10/300 GL
column has been used to determine if inhibition of RNR by
ClFTP or ClFDP is associated with an altered quaternary structure (Table 1 and Fig. S7 A–Q). Experiments were carried out
with various combinations of α or holoenzyme, effectors (dATP,
dGTP, and ATP), and ClF nucleotides. The elution buffer contained 150 mM NaCl and either 0.5 mM ATP (entries 1–14),
20 μM dATP (entry 20), or no nucleotides (entries 15–19). Without effectors, α is a monomer (19) with retention time (RT) of
26 min (Table 1, entry 18, and Fig. S7M). β migrates as a dimer
with a RT of 26 min. When α alone or holoenzyme effectors
(dGTP or ATP) were injected onto the column and eluted with
ATP, the protein eluted in broad peaks over approximately 10 min
with variable RTs and maximum absorbance from 22–25 min,
indicative of interconverting species (Table 1 and Fig. S7 F–I).
The relationship between quaternary structure and holoenzyme activity has also been investigated, and a model based on
kinetic analysis under conditions similar to dynamic light scattering and AUC measurements has been proposed (16, 17). There is
a consensus that α2β2 is the minimal complex required for active
RNR (16–19). However, additional complexes found under physiological conditions (α6β6, α6β2) have also been suggested to be
active (16, 17). Finally, the importance of quaternary structure to
the inhibited state of hRNR inactivated with F2 CDP has also
been reported (4). Using SEC, α6β6 was suggested to be the inactivated state.
In light of these studies and our results with ClFDP and ClFTP,
the quaternary structure again moves front and center as a unique
mode of RNR inhibition. Both nucleotides cause aggregation to
α6 and to α6βn (n ≤ 2) when 1∶1 α∶β is present. In the case of
ClFDP, the initial slow binding and associated conformational
change lead to an inactive state with a dissociative t1∕2 of
23 min. In the case of ClFTP, the rapidly formed inactivated complex undergoes a conformational transition(s) on a minute time
scale to recover approximately 50% of the activity. The molecular
and structural basis of these changes are actively being investigated by cryoelectron microscopy, crystallography, and additional
biochemical experiments.
What are the implications of these studies for hRNR as a therapeutic target? Although it is still early, a few observations can be
made. First the ratio of the ClF metabolites and their stabilities
will likely play a key role in the duration and extent of RNR inhibition. In peripheral mononuclear cells isolated from chronic
lymphocytic leukemia and acute myeloid leukemia patients,
and in cultured CEM cells under certain conditions, ClFMP
levels are elevated and present for extended times and can serve
1. Bonate PL, et al. (2006) Discovery and development of clofarabine: A nucleoside analogue for treating cancer. Nat Rev Drug Discovery 5:855–863.
2. Zhenchuk A, Lotfi K, Juliusson G, Albertioni F (2009) Mechanisms of anti-cancer action
and pharmacology of clofarabine. Biochem Pharmacol 78:1351–1359.
3. Gandhi V, Plunkett W (2006) Clofarabine: Mechanisms of action, pharmacology and
clinical investigations. Cancer Drug Discovery and Development: Deoxynucleoside
Analogs in Cancer Therapy, ed GJ Peters (Humana Press, Totowa, NJ), pp 153–171.
4. Wang J, Lohman GJS, Stubbe J (2007) Enhanced subunit interactions with gemcitabine-5′-diphosphate inhibit ribonucleotide reductases. Proc Natl Acad Sci USA
104:14324–14329 and references cited therein.
5. Carson DA, et al. (1992) Oral antilymphocyte activity and induction of apoptosis by
2-chloro-2′-arabino-fluoro-2′-deoxyadenosine. Proc Natl Acad Sci USA 89:2970–2974.
6. Lofti K, et al. (1999) Biochemical pharmacology and resistance to 2-chloro-2′-arabinofluoro-2′-deoxyadenosine, a novel analogue of cladribine in human leukemic cells.
Clin Cancer Res 5:2438–2444.
7. Parker WB, et al. (1991) Effects of 2-chloro-9-(2′-deoxy-2′-fluoro-β-D-arabinofuranosyl)
adenine on K562 cellular metabolism and the inhibition of human ribonucleotide
reductase and DNA polymerases by its 5′-triphosphate. Cancer Res 51:2386–2394.
8. Xie KC, Plunkett W (1995) Metabolism and actions of 2-chloro-9-(2′-deoxy-2′-fluoroD-arabinofuranosyl)adenine in human lymphoblastoid cells. Cancer Res 55:2847–2852.
9. Xie KC, Plunkett W (1996) Deoxynucleotide pool depletion and sustained inhibition of
ribonucleotide reductase and DNA synthesis after treatment of human lymphoblastoidcells with 2-chloro-(2′-deoxy-fluoro-D-arabinofuranosyl) adenine. Cancer Res
56:3030–3037.
10. Wang J, Lohman GJS, Stubbe J (2009) Mechanism of inactivation of human
ribonucleotide reductase with p53R2 by gemcitabine 5′-diphosphate. Biochemistry
48:11612–11621.
11. Lohman GJS, Stubbe J (2010) Inactivation of Lactobacillus leichmannii ribonucleotide
reductase by 2′,2′-difluoro-2′-deoxycytidine 5′-triphosphate: Covalent modification.
Biochemistry 49:1404–1417.
12. Harris G, Ashley GW, Robins MJ, Tolman RL, Stubbe J (1987) 2′-Deoxy-2′-halonucleotides as alternate substrates and mechanism-based inactivators of Lactobacillus leichmannii ribonucleotide reductase. Biochemistry 26:1895–1902.
13. Stubbe J, van der Donk WA (1998) Protein radicals in enzyme catalysis. Chem Rev
98:705–762.
14. Licht SS, Stubbe J (1999) Mechanistic inverstigations of ribonucleotide reductases.
Comprehensive Natural Product Chemistry, ed CD Poulter (Elsevier, Amsterdam),
pp 163–204.
15. Nordlund N, Reichard P (2006) Ribonucleotide reductases. Annu Rev Biochem
75:681–706.
9820 ∣
www.pnas.org/cgi/doi/10.1073/pnas.1013274108
as a reservoir for di- and triphosphate states of the drug (6–9).
Whether the observed characteristics including the relative ratios
of ClFDP∶ClFTP hold for other cell lines will play an important
role in understanding hRNR inhibition.
Our results in vitro indicate that both ClFDP and ClFTP
reversibly inhibit the enzyme and that the α•ClFDP* tight complex once formed prevails for some time (also see Fig. S4D and
SI Text, p. 4). Thus it is possible, given prolonged retention times
of all ClF nucleotides in the cell (8), that even if the triphosphate
is present at much higher concentrations than the diphosphate,
the diphosphate will eventually be converted into an inhibited
state [E • I , Scheme 1 (ii)] with a moderately long half-life.
Our studies implicate the importance of ClFDP in RNR inhibition. They further indicate the importance of quaternary structure of α alone for targeting inhibitors. Detailed structural
analysis of the different inhibited states and screening methods
with fluorescent probes to determine formation of these states
(30) could lead to yet unidentified compounds that alter RNR’s
quaternary structure to cause its inhibition.
Materials and Methods
See SI Text for details. Syntheses of ([8-3 H]-) ClFDP and ClFTP were carried
out as described in SI Text. In all assays, dCDP/dADP production was analyzed
according to the method of Steeper and Steuart (for dC) (31) or Cory et al.
(for dA) (32), respectively, subsequent to dephosphorylation using alkaline
phosphatase.
ACKNOWLEDGMENTS. We thank the reviewers for insightful comments.
This research was supported by National Institutes of Health Grant GM29595
(to J.S.) and Damon Runyon Cancer Research Fellowship DRG2015-09
(to Y.A.).
16. Kashlan OB, Cooperman BS (2003) Comprehensive model for allosteric regulation of
mammalian ribonucleotide reductase: Refinements and consequences. Biochemistry
42:1696–1706.
17. Rofougaran R, Vodnala M, Hofer A (2006) Enzymatically active mammalian ribonucleotide reductase exists primarily as an α6β2 octamer. J Biol Chem 281:27705–27711.
18. Ingemarson R, Thelander L (1996) A kinetic study on the influence of nucleoside
triphosphate effectors on subunit interaction in mouse ribonucleotide reductase.
Biochemistry 35:8603–8609.
19. Fairman JW, et al. (2011) Structural basis for allosteric regulation of human ribonucleotide reductase by nucleotide-induced oligomerization. Nat Struct Mol Biol
18:316–322.
20. Thelander L (2007) Ribonucleotide reductase and mitochondrial DNA synthesis. Nat
Genet 39:703–704 and references cited therein.
21. Stubbe J, Nocera DG, Yee CS, Chang MCY (2003) Radical initiation in the class I
ribonucleotide reductase: Long-range proton-coupled electron transfer. Chem Rev
103:2167–2202.
22. Reichard P, Eliasson R, Ingemarson R, Thelander L (2000) Cross-talk between the
allosteric effector-binding sites in mouse ribonucleotide reductase. J Biol Chem
275:33021–33026.
23. Morrison JF, Walsh CT (1988) The behaviour and significance of slow-binding enzyme
inhibitors. Adv Enzymol Relat Areas Mol Biol 61:201–301.
24. Copeland RA (2005) Evaluation of Enzyme Inhibitors in Drug Discovery: A Guide for
Medicinal Chemists and Pharmacologists (Wiley-Interscience, New York).
25. Copeland RA (2000) Enzymes: A Practical Introduction to Structure, Mechanism, and
Data Analysis (Wiley, New York), 2nd Ed.
26. Cook PF, Cleland WW (2007) Enzyme Kinetics and Mechanism (Garland Science,
London), pp 195–204.
27. Cory JG, et al. (1985) Nucleoside 5′-diphosphates as effectors of mammalian ribonucleotide reductase. J Biol Chem 260:12001–12007.
28. Chimploy K, Mathews CK (2001) Mouse ribonucleotide reductase control: Influence
of substrate binding upon interactions with allosteric effectors. J Biol Chem
276:7093–7100.
29. Nakamura CE, Abeles RH (1985) Mode of interaction of β-hydroxy-β-methylglutaryl
coenzyme A reductase with strong binding inhibitors: Compactin and related
compounds. Biochemistry 24:1364–1376.
30. Hassan AQ, Wang Y, Stubbe J (2008) Methodology to probe subunit interactions in
ribonucleotide reductases. Biochemistry 47:13046–13055.
31. Steeper JR, Steuart CC (1970) A rapid assay for CDP reductase activity in mammalian
cell extracts. Anal Biochem 34:123–130.
32. Cory JG, Russell SA, Mansell MM (1973) A convenient assay for ADP reductase activity
using Dowex-1-borate columns. Anal Biochem 55:449–456.
Aye and Stubbe