Mechanistic Studies of Semicarbazone Triapine Targeting Human Ribonucleotide Reductase in Vitro and in Mammalian Cells

Background: Diferric-tyrosyl radical [(FeIII2-Y·)(FeIII2)] cofactor-bearing subunit (β2) of ribonucleotide reductase is targeted by a Phase-II cancer drug, Triapine (3-AP). Results: Y· loss precedes iron loss without reactive oxygen species formation. Conclusion: Fe(II)-(3-AP) inhibits β2 catalytically resulting in iron-loaded β2 with a reduced Y·. Significance: Susceptibility of β2 to inhibition via Y· reduction by metal complexes implicates a new avenue to develop RNR inhibitors. Triapine® (3-aminopyridine-2-carboxaldehyde thiosemicarbazone (3-AP)) is a drug in Phase II trials. One of its established cellular targets is the β2 subunit of ribonucleotide reductase that requires a diferric-tyrosyl-radical [(FeIII2-Y·)(FeIII2)] cofactor for de novo DNA biosynthesis. Several mechanisms for 3-AP inhibition of β2 have been proposed; one involves direct iron chelation from β2, whereas a second involves Y· destruction by reactive oxygen species formed in situ in the presence of O2 and reductant by Fe(II)-(3-AP). Inactivation of β2 can thus arise from cofactor destruction by loss of iron or Y·. In vitro kinetic data on the rates of 55Fe and Y· loss from [(55FeIII2-Y·)(55FeIII2)]-β2 under aerobic and anaerobic conditions reveal that Y· loss alone is sufficient for rapid β2 inactivation. OxyblotTM and mass spectrometric analyses of trypsin-digested inhibited β2, and lack of Y· loss from H2O2 and O2̇̄ treatment together preclude reactive oxygen species involvement in Y· loss. Three mammalian cell lines treated with 5 μm 3-AP reveal Y· loss and β2 inactivation within 30-min of 3-AP-exposure, analyzed by whole-cell EPR and lysate assays, respectively. Selective degradation of apo- over [(FeIII2-Y·)(FeIII2)]-β2 in lysates, similar iron-content in β2 immunoprecipitated from 3-AP-treated and untreated [55Fe]-prelabeled cells, and prolonged (12 h) stability of the inhibited β2 are most consistent with Y· loss being the predominant mode of inhibition, with β2 remaining iron-loaded and stable. A model consistent with in vitro and cell-based biochemical studies is presented in which Fe(II)-(3-AP), which can be cycled with reductant, directly reduces Y· of the [(FeIII2-Y·)(FeIII2)] cofactor of β2.

The diverse properties of 3-AP have resulted in a number of models by which it inhibits RNR. In one model free 3-AP is proposed to chelate the Fe(III) directly from the [(Fe III 2 -Y⅐)(Fe III 2 )] within ␤ 2 (14,28,29), resulting in RNR inactivation. In a second model 3-AP is proposed to chelate iron from the intracellular iron pool(s) (18,20) that could interfere with the essential [(Fe III 2 -Y⅐)(Fe III 2 )] assembly on ␤ 2 (30). A third model, which is the one currently favored in the literature, is that Fe(III)-(3-AP) is reduced to Fe(II)-(3-AP) by endogenous reductants, which in turn reacts with O 2 and produces ROS that inactivate RNR (21,(27)(28)(29). A direct interaction of Y⅐ with O 2 .
has also been proposed to lead to ␤ 2 inhibition observed in vitro (27)(28)(29). Although depletion of the labile iron pools and ROS generation are likely involved in late-stage cytotoxic pathways such as induction of apoptosis (21,23,24), we will argue that neither mechanism is likely to be important in RNR-specific inhibition. Previous pharmacological studies by Sartorelli and co-workers (13) and Keppler and co-workers (22) that examined the 3-AP-induced late-stage cytotoxicity in a number of cell lines over 2-4 days noted that blockage of DNA synthesis is induced within the initial hours of 3-AP incubation. Studies by Richardson and co-workers (18) have indicated that cellular iron uptake and efflux are minimally perturbed in 3-AP-treated cultured cells in the first hours of exposure. These interesting findings prompted us to undertake a detailed investigation to understand the mechanism of 3-AP-promoted inhibition of ␤ 2 in vitro and in cell culture in the early stages of inhibitor treatment where cell viability remains high, cell cycle is not perturbed, and downstream cytotoxicity is not yet apparent.
Our studies in vitro have investigated the effect of 3-AP and its iron complexes on RNR activity, depletion of its essential Y⅐, and iron loss from the [(Fe III 2 -Y⅐)(Fe III 2 )] cluster of ␤ 2 alone and in the active holo-complex [(␣ 2 ) m (␤ 2 ] in a non-cycling or cycling state. The relative rates of Y⅐ loss and iron loss have been measured and vary dependent upon the states of ␤ 2 (␤ 2 alone, cycling-or non-cycling holo-complex). In all cases, however, Y⅐ loss precedes iron loss, suggesting that ␤ 2 -specific inhibition can be explained by the direct reduction of Y⅐. These data implicate Fe(II)-(3-AP) as the reductant that can rapidly reduce Y⅐. The observations that Y⅐ loss occurs in the presence and, importantly, in the absence of O 2 and that Y⅐ levels are unaf-fected by O 2 . or H 2 O 2 and the failure to detect oxidative modifications by additional studies using Oxyblot TM technology and mass spectrometry on inhibited ␤ 2 rule out ROS as the basis for Y⅐ reduction. We have also investigated the effects of 3-AP in K562, COS-1, and hydroxyurea (HU)-resistant TA3 cells on ␤ 2 activity in cell lysates, 55 Fe content of ␤ 2 immunoprecipitated from 55 Fe-labeled cells, the stability of the inhibited ␤ 2 , and the Y⅐ levels in intact cells. Furthermore, Oxyblot TM technology of the whole protein pool failed to reveal differences between 3-APtreated and untreated cells. Together with in vitro data, the rapid and potent RNR inhibition in cells within 30 min of 3-AP treatment is most consistent with the formation of Fe(II)-(3-AP) from free 3-AP with intracellular Fe, which then engages in the direct reduction of the essential Y⅐ of ␤ 2 . The inhibition occurs in the period where cell viability remains high, and the activity of ␣ 2 and other abundant iron-dependent enzymes such as aconitase, a sensitive indicator of oxidative stress, are maintained. These findings further underscore the specificity of 3-AP-promoted ␤ 2 -specific inhibition and downplay nonspecific indirect models discussed above. They provide initial insight into understanding how inhibition of ␤ 2 is involved in the cytotoxic effects of 3-AP.

EXPERIMENTAL PROCEDURES
In Vitro Assays on hRNR-All in vitro studies were performed on hRNR subunits recombinantly expressed and purified from Escherichia coli (10,11). Specific activities for CDP reduction were 700 -890 nmol min Ϫ1 mg Ϫ1 for ␣ and 3000 -4100 nmol min Ϫ1 mg Ϫ1 for ␤ containing 0.9 -1.2 Y⅐/␤ 2 and 3.6 iron/␤ 2 . The Y⅐ in the [(Fe III 2 -Y⅐)(Fe III 2 )] cluster of ␤ 2 has a t1 ⁄ 2 ϳ 25 min Holo-␤ 2 upon association with (␣ 2 ) m constitutes hRNR that catalyzes nucleoside diphosphate (NDP) reduction to dNDPs, one of the key upstream processes that dictates dNTP pool homeostasis. Direct inhibition of ␤ 2 can arise from either loss of Y⅐ (Met-␤ 2 formation) or Fe III 2 -Y⅐ (apo␤2 formation). Iron-chelating small molecules such as Triapine (3-AP) may also cause indirect inhibition by depletion of intracellular labile iron pool or by Fe(II)-(3-AP)-catalyzed generation of ROS in the presence of O 2 . Blue inset, this study in vitro and in live mammalian cells suggests that ␤ 2 -specific inhibition involves reduction of Y⅐ elicited by the active reductant, Fe(II)-(3-AP), catalytically recyclable by intracellular reductants, e.g. GSH. This inhibition results in iron-loaded-␤ 2 with Y⅐ reduced, inferred in our model as Met-␤ 2 without the need to invoke ROS. at 37°C in 50 mM Hepes (pH 7.6), 100 mM NaCl. All inhibition data, unless otherwise noted, have thus been adjusted for this intrinsic instability. The presence of 5 mM DTT in the storage buffer is necessary to preserve ␣ 2 activity; thus in all experiments with holo-complex, 85 M DTT is present. All experiments were carried out at physiological concentrations of subunits (0.5-2.5 M (monomer) in 3T6 (6,(31)(32)(33), COS-1, HeLa, and NIH-3T3 cells (12)), except the EPR experiments where sensitivity limits require 5 M ␤ 2 .
[( 55 Fe III 2 -Y⅐)( 55 Fe III 2 )] Cluster Assembly in ␤ 2 -Labeling of the active-site diiron center with 55 Fe and subsequent isolation and characterization are detailed in supplemental Materials and Methods. This procedure typically afforded 55 Fe-labeled ␤ 2 with 1.0 -1.2 Y⅐/␤ 2 (quantitated by EPR). The ferrozine assay (34) gave 3.2-3.6 iron/␤ 2 . Non-specifically surface-bound S ϭ 5/2 Fe III was quantitated by EPR analysis of the same samples at 77 K involving double integration of the g ϭ 4.3 signal relative to Fe III -EDTA standard as detailed previously (35). It was less than 5% (Ͻ0.18 iron/␤ 2 ) of the total iron. Specific activity of ␤ 2 reconstituted using an identical procedure to that described above with unlabeled FeCl 3 was 3000 -4100 nmol min Ϫ1 mg Ϫ1 of ␤. Reconstitution and active-site labeling processes were carried out on untagged and His 6 -tag ␤ 2 (10) and showed that the tag does not affect activity or the amount of surface-bound iron. All the in vitro experiments thus used tagged ␤ 2 . All in vitro experimental procedures are described in detail in the supplemental Materials and Methods.
Studies in Cultured Cells-COS-1, K562, and HU-resistant TA3 cells were selected as representative mammalian cell lines for the following reasons. COS-1 cells typically yield a relatively large quantity of total protein, and the endogenous ␤ 2 activity per mg of total protein is 1.5-3-fold higher than that typically obtained from HeLa and NIH-3T3. K562 was chosen, as previous studies reveal endogenous levels of Y⅐ can be detected by whole cell EPR (36). TA3 cells further support the data from K562 due to enhanced levels of Y⅐ and also allow study of 3-AP in HU-resistant cells. All in-cell experimental procedures are provided in the supplemental Materials and Methods.

Mechanism of hRNR Inhibition by 3-AP in Vitro-All in vitro
studies to date on 3-AP-induced inhibition of mouse and hRNR have primarily focused on the analysis Y⅐ loss as a measure of enzyme inactivation (27)(28)(29). Because ␤ 2 inhibition can arise from loss of iron, loss of Y⅐, or both, we have measured independently the effects of 3-AP and its iron complexes on the rates of Y⅐ and active site iron loss.
Time-dependent Inhibition Assays-To establish if RNR activity is depleted in a time-dependent manner and whether this activity loss is associated with ␤ 2 and/or ␣ 2 , 0.6 M ␤ 2 (␣ 2 ) was incubated with 0.3, 1, or 5 eq of 3-AP [per ␤(␣)] for the indicated times and diluted into an assay mixture with a 7-fold excess of ␣ 2 (␤ 2 ) and incubated an additional 3 min. The results shown in Fig. 2A indicate that with 1 eq or a 5-fold excess of 3-AP per ␤, Ͼ90% of the activity is lost within 10 and 1 min, respectively ( Fig. 2A). The ␣ 2 subunit activity under the same conditions remains unchanged ( Fig. 2A), validating that 3-AP is a ␤ 2 subunit-specific inhibitor. Incubation of ␤ 2 with 0.3 eq of 3-AP/␤, however, resulted in 50% activity loss at 20 min ( Fig.  2A), greater than expected if 3-AP itself is the inhibitor. This result suggests that 3-AP has access to iron either from the endogenous loss from ␤ 2 as previously reported in mouse (37) or from nonspecific surface-bound iron associated with the in vitro cluster assembly (10,11). 3-AP-accelerated iron loss has previously been proposed to play a key role in ␤ 2 inhibition (14,29). If Fe(II)/Fe(III)-(3-AP) is generated, then in the presence of DTT (85 M) that always accompanies ␣ 2 in the assays, the metal chelate can redox-cycle (19,25,26), potentially causing RNR inhibition. These observations are consistent with previous proposals (27)(28)(29) that an iron-complex(es) of 3-AP is(are) the true inhibiting species. The kinetics of iron loss from ␤ 2 and whether it is accelerated by 3-AP were, therefore, examined.
Quantitative Assessment of the Rate of Active-site Iron Loss in the Presence and Absence of 3-AP- 55 Fe-Labeled ␤ 2 was assembled in vitro from apo␤2, resulting in specific radioactivity of 652 cpm/nmol of iron with 3.6 iron/␤ 2 and specific activity in nucleotide reduction and Y⅐ content identical to our previous reports (10,11). Importantly, EPR analysis showed that the presence of non-specifically surface-bound iron is less than 5% of total iron. Thus 55 Fe is predominantly associated with the active site, and our assays monitor the rate of 55 Fe loss from this site. A time course for the rate of iron release at 37°C in the absence and presence of 3-AP was examined under three different sets of conditions: ␤ 2 alone and ␤ 2 with ␣ 2 in a 1:1 molar ratio (1 M) either in the presence of the allosteric effector ATP (non-cycling holo-complex) or in the presence of the ATP, CDP, thioredoxin (Trx)/thioredoxin reductase/NADPH reducing system (cycling holo-complex). In the absence of 3-AP, the data reveal that the [(Fe III 2 -Y⅐)(Fe III 2 )] exhibits very different intrinsic stabilities in the three cases ( Fig. 2, B-D). When ␤ 2 is alone ( Fig. 2B) or with ␣ 2 in a non-cycling state (Fig. 2C), 20% of the 55 Fe is lost over 20 min. Because non-specifically bound 55 Fe is Ͻ5% that of total iron, the loss detected is associated with loss from metallo-cofactor. The data additionally demonstrate that the intrinsic iron loss is not associated with DTT. DTT is absent in the ␤ 2 alone experiment but present at 85 M in the non-cycling holo-complex (compare, Fig. 2, B and C). DTT is required to preserve ␣ 2 stability and is thus carried over from the ␣ 2 storage solution in all in vitro studies on the holocomplex. In contrast, 90% of the iron is spontaneously lost in 20 min under cycling conditions (Fig. 2D). The observed intrinsic lability in the cycling holo-complex is consistent with previous reports on 59 Fe release from the cycling mouse ␤ 2 in the presence of DTT (ϳ60% loss in 30 min) (37). Our data reveal that the intrinsic lability of the cofactor is DTT-independent (Fig. 2, B and C) and that the cycling state of holo-hRNR provides an additional mechanism for iron loss. Identical experiments to those described above were then carried out in the presence of 3-AP. In the case of ␤ 2 alone, a 5-eq excess of 3-AP over ␤ (Fig. 2B) only elicits 20% iron loss in 20 min, similar to the results in its absence (Fig. 2B). This provides compelling evidence that the free ligand does not accelerate active site iron loss on ␤ 2 alone. In contrast, with the addition of ␣ 2 , under conditions that result in non-cycling-or cycling-holo-complex, 3-AP greatly enhances the rate of iron release, leading to almost complete loss in 20 min with a t1 ⁄ 2 of 4.8 min in both cases (Fig. 2, C and D). The observed increase in loss of iron from the holo-complex in the presence of 3-AP likely implies that formation of the holo-complex may lead to a state(s) that allows the free 3-AP to access iron from the active site, a process not feasible in the case of ␤ 2 alone (Fig. 2B). Alternatively, the presence of reductant DTT (carried over with ␣ 2 ), although having no effect on the intrinsic lability of the cofactor, could redox-cycle the in situ-assembled Fe(III)-(3-AP) such that the diferric center in holo-complex is labilized by reduction. Future experiments should provide a mechanistic explanation of these results. Our focus at this time was to determine whether the varied rates and amounts of iron loss observed in the three states of ␤ 2 studied, and whether Y⅐ loss or both are responsible for ␤ 2 inhibition.

Rate of Y⅐ Loss from [(Fe III 2 -Y⅐)(Fe III
2 )]-␤ 2 -EPR studies that monitor Y⅐ were carried out first on ␤ 2 alone under a variety of conditions: with 3-AP alone, with Fe(III)-(3-AP), with Fe(II)-(3-AP) Ϯ O 2 , and with Fe(III)-(3-AP) and GSH Ϯ O 2 (Fig. 2E). The Y⅐ in hRNR-␤ 2 is inherently unstable; t1 ⁄ 2 of 25 min at 37°C (11). Control experiments measuring the effect of Fe(II), Fe(III), GSH, and combinations of these on the rate of Y⅐ loss indicated minimal changes relative to inherent instability. The data shown in Fig. 2E have been adjusted for the respective intrinsic rates of Y⅐ loss.
In the first experiment 5 eq 3-AP/␤ was incubated with ␤ 2 alone, and as shown in Fig. 2E, 80% of Y⅐ is lost within 20 min. The measured rate of Y⅐ loss induced by the 3-AP alone is similar to that reported for mouse ␤ 2 in the absence of DTT (29). However, comparison of the rate of Y⅐ loss with the rate of 55 Fe loss for ␤ 2 alone (Fig. 2B) provides the first quantitative evidence that 3-AP effectively targets Y⅐ during a time frame in which its impact on the diferric center remains minimal (Fig. 2, B versus E). Our in vitro data on ␤ 2 alone thus suggest that Y⅐ quenching is the principle mechanism of enzyme inhibition and leads to iron-loaded ␤ 2 .
To further investigate if Y⅐ quenching precedes iron loss in non-cycling holo-complex in the presence of 3-AP, similar EPR analyses were performed and compared with the rates of 55 Fe loss (Fig. 2, C and E). Y⅐ loss was 80% complete in 1 min (supplemental Fig. S1B), whereas 40% of total iron was lost in the same timeframe (Fig. 2C). Whether the mechanism of iron loss is coupled to Y⅐ loss in this short time frame remains to be estab-  OCTOBER 12, 2012 • VOLUME 287 • NUMBER 42

JOURNAL OF BIOLOGICAL CHEMISTRY 35771
lished. However, if a single mechanism is at play, iron is lost at a rate 5ϫ slower than Y⅐ over the 20-min duration of the assay.

Rapid Y⅐ Reduction Mediated by Fe(II)-(3-AP)-
In an effort to understand the form of 3-AP responsible for inhibition of ␤ 2 , loss of Y⅐ was also examined in the presence of Fe(III)-(3-AP), Fe(II)-(3-AP) Ϯ O 2 , and with Fe(III)-(3-AP) and GSH Ϯ O 2 (Fig. 2E). With 5 eq of preassembled Fe(III)-(3-AP) (per ␤), the rate of Y⅐ loss was similar to the case with 5 eq of free 3-AP/␤ (Fig. 2E). In contrast, when preassembled Fe(II)-(3-AP) was examined, Y⅐ destruction was complete within the first time point. The same rapid Y⅐ loss with a t1 ⁄ 2 of Ͻ0.5 min was observed when the experiment was repeated under strictly anaerobic conditions (Fig.  2E). These data suggest that Fe(II)-(3-AP) is the active species and that ROS are unlikely to be involved.
To provide additional support for proposed role of Fe(II)-(3-AP), the EPR experiment with preassembled Fe(III)-(3-AP) complex and ␤ 2 alone was replicated in an anaerobic chamber in the presence of 5 mM GSH. GSH was chosen as a reductant due to its physiological role and abundance, and the reported range of redox potentials suggests its ability to reduce Fe(III)-(3-AP) (38). The rapid rate of Y⅐ loss with GSH and in the absence of O 2 (Fig. 2E) is consistent with Fe(II)-(3-AP) being the active inhibitor. Importantly, rapid Y⅐ loss is also achieved with substoichiometric amounts of Fe(III)-(3-AP) (0.3 eq/␤) when GSH is present (Fig. 2E). These results together suggest that reduction of Y⅐ can be elicited by Fe(II)-(3-AP) catalytically without ROS formation. The chemical mechanism and kinetics of Y⅐ reduction will be the focus of future studies.  (44). These chemotypes can be probed using Oxyblot TM (Millipore) analysis in which the carbonyls are derivatized to 2,4-dinitrophenyl hydrazones, which are then detected by Western blotting (45). Thus Oxyblots were used to determine if nonspecific oxidative damage could be responsible for ␤ 2 inhibition by 3-AP. ␤ 2 alone and non-cycling-and cycling holo-complexes were incubated with and without 3-AP for 20 min. Proteins in controls and samples with 3-AP were derivatized with 2,4-dinitrophenylhydrazine and subjected to gel electrophoresis followed by Western blot analysis. Protein carbonyls were not detected within ␤ 2 treated with 3-AP under any conditions examined (supplemental Fig. S2A). A positive control was carried out with ␤ 2 treated with excess ascorbate and iron. SDS-PAGE/Western analysis revealed that ␤ (47 kDa monomer) in these conditions underwent a gel shift to higher a molecular weight species, presumably associated with crosslinking during oxidative damage. Interestingly, Oxyblot analysis on the cycling holo-complex indicated that Trx reductase was oxidized, further indicating that the Oxyblot could readily detect oxidation; no oxidation, however, was associated with ␤ 2 (supplemental Fig. S2A). Omission of 2,4-dinitrophenylhydrazide resulted in complete absence of all the bands. Given the reported sensitivity (fM) of the Oxyblot method, the data imply that Fe(II)-(3-AP)-catalyzed protein oxidation is not involved in ␤ 2 inhibition.
Trypsin  Table S1A) and hydroxylation (supplemental Table S1B). Qualitatively, there was no significant difference between these three samples. The modifications detected are likely caused by sample handling during protein isolation, gel electrophoresis, or the in-gel digestion process. The Oxyblot and mass spectrometric data suggest inhibition of ␤ 2 by 3-AP does not result in significant modification of ␤ 2 primary structure.
Mechanism of 3-AP Inhibition in Mammalian Cells-Our attention turned to determine if a similar mechanism occurs in COS-1-, K562-, and HU-resistant TA3 cells exposed to 3-AP.
Loss of ␤ Activity in Cell Lysates from COS-1 and K562 and HU-resistant TA3 Cells Treated with 3-AP-Studies from independent laboratories have reported that cultured cells (HL60 and L1210) exposed to 1-10 M 3-AP for 0.5-4 h result in complete inhibition of DNA synthesis long before loss of cell viability (2-4 days) (13,22). Thus our experiments with COS-1 and K562 cells to study the RNR inhibition mechanism were carried out on cells treated with 3-AP (5-500 M) over 0.5-3 h where the cells maintained high viability (supplemental Figs. S3 and S4). As shown in supplemental Fig. S5, 30 min of exposure of cells to 3-AP at 5 M, similar to the in vivo concentration for ␤ 2 (ϳ2.5 M monomer) (6, 31-33), resulted in no detectable ␤ 2 activity in the resulting lysates. Under the same conditions, ␣ 2 -activity was unaffected. Similar experiments were also carried out with HU-resistant ⌻〈3 cells, previously shown to overexpress ␤ 2 and to have elevated concentrations of Y⅐ by whole-cell EPR analysis (37,46). Consistent with the higher levels of Y⅐, our lysate assay data in these cells showed that ␤ 2 activity is ϳ4-fold higher than that in COS-1 cells (Fig. 3A versus supplemental Fig. S5A). Treatment of TA3 cells with 5 M 3-AP for 30 min still completely abolished ␤ 2 activity (Fig. 3A). Our lysate assay data in all three cell lines thus provided direct evidence that 3-AP promotes rapid loss of RNR activity in live cells and is consistent with previous reports on the early inhibition of DNA synthesis in other cell lines studied (13,22).
Whole Cell EPR Analysis Links ␤ 2 Inhibition to Y⅐ Loss-Whole cell EPR experiments were undertaken using analogous procedures to the previously published reports (36,37,46) S6). In the TA3 cells, the higher level of Y⅐ was reflected in a strong signal that are easily monitored even at 77 K (supplemental Fig. S6A), and the signal was identical to that from recombinant [(Fe III 2 -Y⅐)(Fe III 2 )]-␤ 2 reconstituted in vitro (supplemental Fig. S6B). Treatment of either TA3 or K562 cells with 5 M 3-AP abolished the Y⅐ in as early as 30 min (Fig. 3B and supplemental Fig. S6, C and D). Note that the lower concentration of Y⅐ in ␤ 2 from K562 cells is accompanied by higher background signals (supplemental Fig. S6D) as previously reported (36). Thus whole cell EPR analyses and lysate activity assays together demonstrate that 3-AP targets ␤ 2 in these cells.
Levels of ␤ 2 Are Maintained in 3-AP-treated Cells 12 h Subsequent to Activity Loss-The absence of ␤ 2 activity and loss of Y⅐ could be accompanied by conversion of ␤ 2 to an iron-loaded or apo form (Fig. 1). To address this issue, Western blot analysis was carried out on non-synchronized TA3, K562, and COS-1 cells at 30 min (supplemental Fig. S7). These studies showed that in all three cell lines, ␤ 2 levels were similar to untreated cells ( Fig. 3C and supplemental Fig. S7). In addition, a time course study showed that ␤ 2 expression was unchanged up to a 12-h period of exposure to 5 M 3-AP (Fig. 4B). Cells also maintained their high viability over this time (supplemental Fig. S3).

Analysis of Stability of Apo␤2 and Holo-␤ 2 in Mammalian
Lysates to Proteasomal Degradation-The in vivo data thus far suggest that 3-AP-dependent inhibition is associated with formation of iron-loaded-␤ 2 but with Y⅐ reduced or apo␤2. However, our in vitro analyses suggest that Y⅐ loss alone is sufficient to account for ␤ 2 inhibition (Fig. 2, B-D, and supplemental Fig.  S1). Two approaches were undertaken to determine the cofactor status in inhibited ␤ 2 . The cell cycle-dependent periodicity of mammalian ␤ 2 is regulated by ubiquitin-mediated proteasomal degradation during the passage into mitotic (M) phase (6,47). This knowledge suggests that ␤ 2 can be primed for degradation by the 26 S proteasome, although the cofactor status of the ␤ 2 being degraded during the M phase remains an outstanding interesting question. We thus started our study using a proteasomal degradation assay (48). COS-1 lysates were chosen to allow comparison with Western blot analysis above. Freshly prepared lysates (1 mg/ml) were supplemented with an ATP-regenerating system and incubated with either apo␤2 or [(Fe III 2 -Y⅐)(Fe III 2 )]-␤ 2 . Aliquots of each reaction were removed at the indicated time points from 0 to 60 min and quenched in Laemmli buffer. Western blot analysis revealed that apo␤2 had a t1 ⁄ 2 of 25 min, whereas the holo-␤ 2 level remained unaltered at 1 h (Fig. 4A). To verify that the proteasome was involved in this degradation, bortezomib, a clinically used proteasome inhibitor (49), was added to separate lysates from the same batch as those shown to degrade apo␤2, and the experiment was repeated (Fig.  4A). No degradation was observed, confirming that apo␤2 is targeted by the proteasome.
Stability of Inhibited ␤ 2 Over Prolonged Periods-At this point it is unclear whether ␤ 2 levels (Fig. 4B) are unchanged or they appear unchanged due to up-regulation of ␤ 2 expression resulting from 3-AP treatment compensated for by increased ␤ 2 degradation (consistent with apo␤2 formation). Up-regulation in expression of target enzymes upon inhibitor treatment has been reported in mammalian cells (50 and 68) and in many cases and can occur on a 3-h time scale (51). Cycloheximide (CHX), an inhibitor of new protein synthesis (52), was used with 3-AP-treated and untreated cells to gain mechanistic insight. Because ␤ 2 has a t1 ⁄ 2 ϳ 3 h in mammalian cells (6, 46, 53),

. Selective degradation of apo␤2 over holo-␤ 2 by the proteasome in COS-1 lysates (A) and the prolonged stability of inhibited ␤ 2 in 3-AP-treated COS-1 cells (B).
A, apo␤2 is selectively primed for degradation with t1 ⁄2 ϭ 25 min, but holo-␤ 2 persists. Specific proteasome inhibiting drug, bortezomib (Bo), prevents apo␤2 degradation. B, ␤ 2 protein levels were maintained over the 12-h treatment of COS-1 cells with 5 M 3-AP. GAPDH was used as the loading control. High cell viability is maintained throughout this period (supplemental Fig. S3A). Activity depletion occurs within 30 min of exposure to 3-AP (supplemental Fig. S5A). Shown here in each diagram are representative Western blots from duplicate sets of experiments.
we opted to compare the effect of 3-AP on the levels of ␤ 2 expression in the presence of CHX. Non-synchronized COS-1 cells in logarithmic phase were treated with CHX (0.2 mg/ml) or DMSO for 0.5 h, at which time the media in half of the CHX-treated samples as well as half of DMSO-treated samples were replaced with media that contained 5 M 3-AP and CHX (0.2 mg/ml) (Fig. 5A, lane c) The data from this set were labeled (3-AP ϩ CHX; Fig. 5B). The other half of the samples was retreated with CHX (Fig. 5A, lane  a) or DMSO only (lane b). The data from this set were labeled CHX (Fig. 5B). All samples were then incubated for an additional 3 h. This time frame also limits the off-target effects from CHX-induced cytotoxicity (54). Western blot analyses showed that the addition of CHX, as expected, decreased the steadystate level of ␤ 2 by ϳ60% over 3 h (Fig. 5, A, compare lanes a and  b, and B, CHX), consistent with the reported t1 ⁄ 2 of ␤ 2 in TA3 and 3T6 cells (6,46,53). When this experiment was repeated with 5 M 3-AP (Fig. 5, A, compare lanes c and d, and B, 3-AP ϩ CHX), a similar ϳ60% decrease in ␤ 2 levels was also observed. Thus ␤ 2 stability was unaffected by 3-AP treatment. Given that apo␤2 appears to be unstable in the lysate degradation assay, ironloaded ␤ 2 is the most probable species remaining in the 3-APtreated cells. 55

Fe-Pulse Labeling and Immunoprecipitation of ␤ 2 Demonstrates No Change in Iron Loss Ϯ 3-AP-treated Cells-
The second approach undertaken to examine the state of inhibited ␤ 2 involves immunoprecipitation from 3-AP-treated and -untreated cells that have been prelabeled with 55 Fe. The labile nature of the diferric center revealed by our in vitro studies on 55 Fe-labeled ␤ 2 (Fig. 2, B-D) mandates rapid isolation of ␤ 2 . An optimized protocol typically affords ϳ50 ng of homogeneous ␤ from ϳ80 ϫ 10 6 K562 cells (Fig. 6A). Pulse-labeling was performed according to published protocols (55) using in vitro 55 Fe-labeled human transferrin (Tf) (2.4 ϫ 10 5 cpm/nmol of iron with 2 iron/Tf), which was added (at 25 g/ml) to logarithmically growing K562 cells. Subsequent to 18 h of incubation, the 55 Fe-labeled Tf-containing media was replaced with fresh media, and the cells were treated with 5 M 3-AP for 0.5 h or DMSO (control). The cells were then harvested, and the relative levels of 55 Fe-labeled ␤ 2 between 3-AP-treated and control cells were determined using immunoprecipitation. The results are shown in Fig. 6. Importantly, the observed values were appreciably above the non-specifically bound background values obtained in the replica pulldown experiments in which the antibody was omitted, but the resin maintained (Fig. 6, A, lane  d, and B, lane c). As an additional control, the lysate from the untreated, labeled cells was treated with 10 mM HU during the antibody incubation step, and the immunoprecipitation process was repeated. HU is known to labilize the diiron center from the mouse and hRNR ␤ 2 in vitro (56) 3 and in mammalian lysates (57). Consistent with these reports, only background levels of iron are associated with ␤ 2 when HU is added (Fig. 6B,  lane d). These data in addition to the degradation studies suggest that the iron of ␤ 2 is largely retained during the 30 min of exposure of cells to 5 M 3-AP, although Y⅐ is reduced.    55 Fe-pulse-labeled K562 cells. A, rapid immunoprecipitation was performed with anti-␤ 2 antibody (Abcam) conjugated to Protein G-Dyna beads (Invitrogen). a, immunoprecipitated human ␤ 2 band (*, monomer, 45 kDa) was confirmed with Western blotting using human ␤ 2 -specific antibody (Ab57653). b, M r ladders (Bio-Rad). c, PVDF membrane in a was stained with Coomassie Brilliant Blue. * indicates immunoprecipitated ␤ monomer. d, control as in a except buffer replaced Ab57653 but Protein G-Dyna beads were kept. B, shown is the amount of radioactivity associated with isolated ␤ 2 subsequent to 30-min treatment of the prelabeled cells with 5 M inhibitor. Shown are the results from 3-AP-untreated (a) and -treated cells (b). c, treatment was identical to a, except anti-␤ 2 antibody was omitted but the beads were retained. d, treatment was a except 10 mM HU was added to the lysate. Error bars are S.D. over duplicate sets of experiments.
An Oxyblot of Lysates of 3-AP-treated Cells Shows the Absence of Protein Carbonyls-As shown above in vitro (supplemental Fig. S2A), further validation of the absence of oxidative carbonylation was probed in whole cell lysates resulting from 3-AP-treated cells and a comparison was made with untreated controls. No change in the extent of carbonylation was observed in any of the cell lines studied (supplemental Fig.  S2B), suggesting that 3-AP treatment of the mammalian cells over this time period does not induce oxidative damage.
3-AP Is Specific for hRNR over Other Iron/Sulfur-requiring Enzymes: Lack of Inhibition of Aconitase-Given the recent report on the interconnections between iron/sulfur centers, non-heme iron and heme cofactor biogenesis, and iron homeostasis (58,59), we examined whether 3-AP was selective for ␤ 2 or whether it might affect the activity of other iron-requiring enzymes. Aconitase that contains an essential [4Fe4S] cluster (60) was chosen as a representative enzyme, as its activity is readily assayed in lysates, the cluster is very sensitive to ROS, and one iron is readily lost from the cluster (61). The proposed chelation property of 3-AP could thus affect the availability of labile iron pools from which the iron/sulfur cluster is biosynthesized and repaired (58,62). Thus, aconitase is a sensitive probe of several mechanisms postulated for 3-AP. However, subsequent to 3 h of treatment of COS-1 cells with 5 M 3-AP, no changes in aconitase activity relative to the control were observed. These observations are consistent with other data in this study which indicate that 3-AP plays an active role in RNR inhibition. These results further show that the action of 3-AP is ␤ 2 -specific.

DISCUSSION
Phase II clinical trials on 3-AP have sparked a resurgence of interest in the mechanisms by which 3-AP and other semicarbazones result in cell cytotoxicity (16,17). The studies are complex, as 3-AP and similar species are typically excellent iron chelators (19,20,25,26), and the reduction potential of these complexes allows redox cycling under physiological conditions (19,25,26), which can result in production of ROS (21,(27)(28)(29). 3-AP, in contrast to other iron chelators, does not appear to sequester iron effectively in the cell (18); thus, it is unlikely to be involved in disruption of biosynthesis and repair of many metallo-cofactor-requiring proteins (30,62,63). Early studies in cultured cells established that their treatment with 3-AP resulted in rapid DNA inhibition within the initial hours (13,22) and that this effect was related to loss of activity of RNR known to require a [(Fe III 2 -Y⅐)(Fe III 2 )] cofactor (13,14). This cofactor is common to all class Ia RNRs (holo-␤ 2 , Fig. 1) (3,4,30), although the in vitro the half-life of the Y⅐ varies from days (E. coli) to minutes (Pseudomonas aeruginosa and human) (11,64). In addition, the mouse ␤ 2 under conditions in which deoxynucleoside diphosphates are produced has been reported to spontaneously lose iron (ϳ60% loss in 30 min at 37°C) (37). Because the [(Fe III 2 -Y⅐)(Fe III 2 )] cofactor is essential for RNR activity (3,4), loss of either Y⅐ or iron cluster results in inhibition.
Studies in cultured cells and xenograft models treated with 3-AP on the other hand have in general focused on assessment of 3-AP-promoted late-stage cytotoxicity (13-15, 18 -24, 28, 29). These data suggest that multiple pathways are operative. Postulated mechanisms include deficiency in DNA lesion repair (15), depletion of intracellular iron storage pools (18,20), ROS-mediated oxidative damage (21,24,(27)(28)(29), and initiation of apoptotic signals (21,23,24). Our data now provide a window into the initial period of 3-AP treatment of cells before loss of regulation at multiple levels impact cell viability. A unified model results from comparison of in vitro and cell culture studies despite many remaining questions that need to be resolved.
In vitro-The currently favored model for the mechanism of RNR inhibition by 3-AP in vitro is that Fe(III)-(3-AP) can be reduced to Fe(II)-(3-AP) by DTT in assay buffers (27)(28)(29) or ␤ 2 itself (29) and generate ROS that reduce the Y⅐. Yen and coworkers (27) studying hRNR ␤ 2 alone by EPR analysis showed complete loss of Y⅐ subsequent to a 30-min incubation with a catalytic amount of Fe(II)-(3-AP) (preassembled in the presence of excess DTT). Under the same conditions, Fe(III)-(3-AP) without DTT resulted in ϳ60% Y⅐ loss, whereas 3-AP alone resulted in ϳ40% Y⅐ loss (see Fig. 4 in Yen and co-workers (27)). These results are similar to our EPR data shown in Fig. 2E. Yen and co-workers (27) further showed using a 5,5-dimethyl-1pyrroline-N-oxide as a radical trap that incubation of Fe(II)-(3-AP) with O 2 generates ROS and that superoxide dismutase and catalase partially prevented Y⅐ loss. From all their data they concluded that O 2 is important in Y⅐ destruction. Their complex experimental protocol, however, involved preincubation of Fe(II)-(3-AP) for 30 min with catalase alone or in combination with superoxide dismutase before addition to ␤ 2 followed by a further 30-min incubation. The experimental design (27) and the observation that DTT, Fe(II), and O 2 can rapidly regenerate Y⅐ (65) suggest that an alternative interpretation of their results is possible; that is, that Y⅐ is regenerated.
Gräslund and co-workers (29) has also reported time-dependent Y⅐ loss on mouse ␤ 2 alone. Their EPR data in the presence of DTT (see Fig. 6 in Gräslund and co-workers (29)) support Fe(II)-(3-AP) as the active species in Y⅐ reduction, consistent with Yen and co-workers (27,28) and our own data. However, as in Yen and co-workers (27,28), Gräslund and co-workers (29) favor the importance of ROS in Y⅐ reduction. This proposal was based on the observation that they observed 100% loss in Y⅐ in 5 min at 37°C with substoichiometric amounts of Fe(III)-(3-AP) relative to ␤ 2 but only 20% loss under anaerobic conditions (see Fig. 6a in Gräslund and co-workers (29)). The source of reducing equivalents essential for reduction of Fe(III)-(3-AP) to Fe(II)-(3-AP) was suggested to be mouse ␤ 2 itself. We note that with Fe(III)-(3-AP) and O 2 , neither our laboratory (Fig. 2E) nor Yen and co-workers (27) see 100% Y⅐ loss, although the possibility remains that hRNR-␤ 2 is distinct from the mouse enzyme. The basis for the differences in the results remains unresolved. Finally, although both the Yen (27,28) and Gräslund laboratories (29) have suggested that 3-AP can bind to ␤ 2 , in the former case based on binding experiments with 3-[ 3 H]AP and the latter case based on molecular docking and computational analysis, in neither case is the proposal compelling.
We propose an alternative model, consistent with our biochemical studies, that accommodates most of the previously reported results. Our kinetic data on Y⅐ and 55 Fe loss from ␤ 2 suggest that rapid inhibition of hRNR-␤ 2 results from reduction of Y⅐ by Fe(II)-(3-AP) by outer sphere electron transfer (Fig. 2, B-E). This mechanism also can account for catalytic cycling of this reductant but does not involve O 2 (Fig. 2E). Our studies further suggest that met-␤ 2 is likely generated (Fig. 2B) and that this state persists for 20 min, the duration of the assay. Given recent studies on the P. aeruginosa ␤ (64), however, we cannot rule out the possibility that Fe II 2 is formed and remains tightly bound. Our additional experiments reveal that presence of ␣ 2 appears to allow 3-AP to remove the iron from the holocomplex, although the rate of iron loss is slower than Y⅐ loss (Fig. 2, C and D, supplemental Fig. S1B). Finally, efforts to detect evidence for damage to the inhibited ␤ 2 using oxyblot methods (supplemental Fig. S2A) and mass spectroscopy analysis of peptides of in-gel-trypsinized ␤ 2 (supplemental Table  S1) failed to reveal protein damage that could account for its inhibition. Thus our data in vitro show that the inhibition of ␤ 2 occurs without the need to invoke ROS formation. We also observe a distinction between the effects of 3-AP on Y⅐ reduction and chelation of active-site iron (Fig. 2, B--E). A detailed mechanism of Fe(II)-(3-AP) reduction of the Y⅐ remains to be established in vitro.
In Vivo-Our studies in cell culture further support our in vitro mechanism for 3-AP-mediated ␤ 2 inhibition that occurs in the initial hours of drug treatment (0.5-4 h) (13,22). Evidence is provided from whole cell EPR methods to monitor Y⅐ (Fig. 3B, supplemental Fig. S6, C and D), 55 Fe-labeled Tf labeling, and immunoprecipitation of the inhibited ␤ 2 to assess iron content (Fig. 6) and oxyblot experiments of the 3-AP treated cells (supplemental Fig. S2B). Our results demonstrate that Y⅐ loss occurs in multiple cell lines after only a 30-min exposure to 3-AP (Fig. 3B, supplemental Fig. S6, C and D). Assays of accompanying cell lysates revealed no detectable RNR activity (Fig.  3A, supplemental Fig. S5) or iron loss in the same time frame (Fig. 6B). Finally, no enhanced oxidation due to 3-AP treatment of the live cells was detected (supplemental Fig. S6B).
In the cell the actual state(s) of ␤ 2 (holo-complex versus ␤ 2 alone) has not been reported. Our studies indicate that the final state of ␤ 2 is iron-loaded with its Y⅐ reduced (Figs. 4A and 6). The inhibited state of ␤ 2 is surprisingly stable even under conditions where new protein synthesis is inhibited (Fig. 5). Our findings suggest that apo␤ 2 is not the end product in 3-APtreated cells. Finally, ␣ 2 is also present in cells, conditions where 3-AP can remove iron from ␤ 2 in a holo-complex based on our in vitro findings (Fig. 2, C and D). Thus, although a portion of ␤ 2 may undergo iron loss, it may be reloaded, giving rise to the observed protein stability in 3-AP-treated cells (Figs. 4B and 5). Preservation of the diiron center in ␤ 2 is consistent with previous studies with SK-N-MC cells treated with 50 M 3-AP for 3 h which revealed that iron uptake from 59 Fe-labeled Tf (ϳ20%) and iron release from 59 Fe pulse-labeled cells (25%) were both minimal relative to the other iron chelators studied (18). Consistent with these results, whole cell EPR analyses of 3-APtreated K562 cells (5 or 500 M) after 0.5, 3, and 12 h showed no change in the size of the intracellular g ϭ 4.3 signals between the treated and untreated cells (data not shown). These observations contrast to the phenotype typically observed when chelation and depletion of intracellular iron pools are in operation (36).
Our studies thus raise interesting questions about the mechanisms by which the diiron cluster of ␤ 2 persists and whether the [(Fe III 2 -Y⅐)(Fe III 2 )]-␤ 2 can be regenerated by a maintenance path-way in vivo, with the Y⅐ continually being re-reduced by cycling Fe(II)-(3-AP). The identification of a small molecule that can negatively intercept the regeneration of active cofactor in living mammalian cells (Fig. 1) may ultimately shed light on the [(Fe III 2 -Y⅐)(Fe III 2 )] cluster biosynthetic and maintenance pathways (30). Our model is that Fe(II)-(3-AP) is the active species involved in ␤ 2 inhibition (Fig. 1). This proposal is in accord with the reported detection of EPR signals (g ϭ 2) in peripheral blood mononuclear cells collected from patients treated with 3-AP (66) within 2 h of treatment, proposed to be iron-and copperbound 3-AP. Given the intracellular prevalence of iron relative to Cu (63) and the stability constants for 3-AP (19,24,25,29), Fe(II)-(3-AP) is the most likely active species. Our in vitro studies with 55 Fe-labeled ␤ 2 suggest the protein itself is capable of providing at least catalytic amounts of iron in all states of ␤ 2 (Fig. 2, B-D); in vivo, however, multiple sources of iron are present. In the continued presence of free ligand in cultured cells, RNR inhibition can persist by the redox cycling capacity of Fe(II)/(III) complex (19,25,26), facilitated by intracellular reductants such as GSH (38). Finally, although our model ( Fig.  1) primarily focuses on targeting ␤ 2 alone, our data in vitro and in non-synchronized cells suggest that the holo-complex is also susceptible to 3-AP-promoted inhibition by direct Y⅐ reduction (Fig. 2, C and D and supplemental Fig. S1B).
Our findings demonstrate that Y⅐ within hRNR ␤ 2 in any state (alone or non-cycling or cycling holo-complex) is highly susceptible to reduction while being stable to a range of reactive small-molecule oxidants. This observation is significant because a pool of ␤ 2 unbound to (␣ 2 ) m almost certainly exists in cells, and our in vitro data show that this form of ␤ 2 is relatively inert to iron chelation, at least by 3-AP. Thus to design more effective semicarbazone anticancer agents, the efficiency of the reduction of hRNR Y⅐ by the semicarbazone Fe(III)/(II) complexes should be considered alongside other pharmacokinetic properties. Furthermore, because Y⅐ is highly susceptible to reduction, alternative reductants/radical quenchers should be investigated as potential ␤ 2 inhibitors (19). Finally, our studies provide a simple series of experiments that can be used to distinguish between ROS effects and Y⅐ reduction both in vitro and in cells, allowing a more thorough evaluation/optimization of other semicarbazones as potential therapeutics.