Glutamate 52-β at the α/β subunit interface of Escherichia coli class Ia ribonucleotide reductase is essential for conformational gating of radical transfer

Ribonucleotide reductases (RNRs) catalyze the conversion of nucleoside diphosphate substrates (S) to deoxynucleotides with allosteric effectors (e) controlling their relative ratios and amounts, crucial for fidelity of DNA replication and repair. Escherichia coli class Ia RNR is composed of α and β subunits that form a transient, active α2β2 complex. The E. coli RNR is rate-limited by S/e-dependent conformational change(s) that trigger the radical initiation step through a pathway of 35 Å across the subunit (α/β) interface. The weak subunit affinity and complex nucleotide-dependent quaternary structures have precluded a molecular understanding of the kinetic gating mechanism(s) of the RNR machinery. Using a docking model of α2β2 created from X-ray structures of α and β and conserved residues from a new subclassification of the E. coli Ia RNR (Iag), we identified and investigated four residues at the α/β interface (Glu350 and Glu52 in β2 and Arg329 and Arg639 in α2) of potential interest in kinetic gating. Mutation of each residue resulted in loss of activity and with the exception of E52Q-β2, weakened subunit affinity. An RNR mutant with 2,3,5-trifluorotyrosine radical (F3Y122•) replacing the stable Tyr122• in WT-β2, a mutation that partly overcomes conformational gating, was placed in the E52Q background. Incubation of this double mutant with His6-α2/S/e resulted in an RNR capable of catalyzing pathway-radical formation (Tyr356•-β2), 0.5 eq of dCDP/F3Y122•, and formation of an α2β2 complex that is isolable in pulldown assays over 2 h. Negative stain EM images with S/e (GDP/TTP) revealed the uniformity of the α2β2 complex formed.

Briefly, initiation of the radical transfer (RT) process is thought to involve proton transfer from the water on Fe 1 in the diferric-Y ⅐ cofactor to Tyr 122 ⅐ and electron transfer from Tyr 356 forming the Tyr 122 phenol ( Fig. 1) (18 -21). Recent RT studies and new Ia RNR subclassifications (22) have helped us to identify conserved residues that could play an important role at the subunit interface of the RNR in conformational gating. The results of these studies are reported herein.
Currently our thinking about the RNR structure is governed by a docking model of ␣2␤2 generated by Eklund and co-workers (3,4) using the crystal structures of ␣2 and ␤2 and their shape complementarity. Their model is supported by four distance measurements (3) made using pulsed electron electron double resonance (PELDOR) spectroscopy and recent biophysical studies including small angle X-ray scattering and single particle electron microscopy (EM) (23)(24)(25). For the most part these methods have taken advantage of RNRs with site-specifically incorporated unnatural amino acids (UAAs) (3). The docking model and the C-terminal 34-amino acid residues of ␤2 served as the starting point for identifying ␣/␤ interface residues.
Two different types of experiments using UAA technology have provided insight about the conformational change(s) effected by binding of S and e to ␣ on the initiation of the rate-limiting conformational gating in ␤ (26,27). One set of experiments used RNR with 3-aminotyrosine (NH 2 Y) site-specifically replacing Tyr 730 in ␣. Incubation of NH 2 Y 730 -␣2, ␤2 with CDP/ATP resulted in loss of Tyr 122 ⅐ in ␤, and formation of a new radical at 730 (NH 2 Y 730 ⅐ ) in ␣2 (24, 26, 28 -30). This oxidation, which occurs only upon binding S and e to ␣, causes an increase in the affinity of the ␣/␤ subunits 25-fold relative to the WT RNR and a decrease in the dissociation rate of the subunits by 10 4 , a process formally involving movement of a single hydrogen atom (24)! A second set of experiments investigating the role of proton-coupled electron transfer at Tyr 356 and the function of the conserved Glu 350 as a proton acceptor of this step (Fig. 1), also provided interesting, unexpected results. Using E350X-␤2 (X ϭ Ala, Asp, or Gln) mutants in WT and mutant backgrounds in which Tyr 122 ⅐ was replaced with tyrosine analogs that are hotter oxidants (3-nitrotyrosine, NO 2 Y 122 ⅐ or 2,3,5-trifluorotyrosine, F 3 Y 122 ⅐ ) (20,31), we found an inability of E350X-␤2 to initiate RT even in the case of the Glu to Asp substitution (18). This result suggested that charged residues might play an important role in gating RT at the interface where the Glu 350 residue resides.
This paper focuses on our efforts to identify additional interface residues using mutagenesis and our ability to site-specifically incorporate UAAs into each subunit. Recently using sequence information, the class Ia RNRs (rnrdb.pfitmap.org/) now designated NrdAg and NrdBg were subcharacterized (22). This information and our current structural understanding of ␣2␤2 resulted in the identification and examination of mutations in four conserved residues: Glu 52 and Glu 350 in ␤2 and Arg 329 and Arg 639 in ␣2 in E. coli RNR. The inactivity of the E. coli mutants established that these residues are essential and the binding studies of ␣/␤ interactions established that with the exception of E52Q, the binding affinities decreased 5-20-fold relative to WT-␣/␤. The tight affinity and inactivity of the E52Q mutant led to further investigation of its properties in the F 3 Y 122 ⅐ background. Unlike WT-␤2, F 3 Y 122 ⅐ -␤2 results in partial uncoupling of the conformational gating that rate limits NDP reduction (20,32) by rapidly producing the Tyr 356 ⅐ (now detectable). It is likely being reduced to the F 3 Y 122 -O Ϫ (phenolate) instead of the phenol (18). Despite the inactivity of E52Q-␤2, the double mutant, E52Q/F 3 Y 122 ⅐ -␤2, when incubated with ␣2/S/e (CDP/ATP or GDP/TTP) resulted in formation of 0.5 eq of the Tyr 356 ⅐ intermediate and in the case of CDP, 0.5 eq of dCDP per F 3 Y 122 ⅐ . Pull-down experiments of the ␣/␤ mixture after 5 min and 2 h using a His 6 -␣, gave a high recovery of a 0.6 -0.8/1.0 ratio of subunits in the ␤/␣ complex. Negative The monomers of ␣2 (PDB code 4R1R) and ␤2 (PDB code 1RIB) are shown in blue and green and red and yellow, respectively. ␣2 was crystallized in the presence of GDP (salmon), TTP (purple), and a peptide corresponding to residues 360 -375 (pink) of ␤2. The ATP cone domain housing the activity site is shown in orange. On the left is the RT pathway between Tyr 122⅐ in ␤2 and Cys 439 in ␣2. Trp 48 is in brackets as there is currently no evidence for its involvement (18). Note Tyr 356 and Glu 350 (not shown) are located in the disordered C-terminal tail of ␤2.
stain electron microscopy (EM) analysis and size exclusion chromatography (SEC) studies revealed that the predominant species is ␣2␤2. The implications of these results on conformational gating and potential structural insight of the active complex are discussed.

Identification of conserved ␣/␤ interface residues, their mutation and assay for activity, and subunit binding affinity
Our recent studies investigating the role of Glu 350 , a conserved residue in the disordered C-terminal tail of ␤2, suggested that this residue was essential for the conformational gating of the RT initiation process (13,18). We therefore looked at other conserved charged residues using the ␣2␤2 docking model, to identify those that might reside at the ␣/␤ interface. Alignment of 80 sequences in the NrdAg/NrdBg subclass revealed that Glu 350 and Glu 52 in ␤, and Arg 639 in ␣ were conserved in 80 of 80 sequences, whereas Arg 329 in ␣ was conserved in 79 of 80. These residues and additional ones, Arg 323 (not conserved) and Arg 735 (76/80) in ␣, became candidates for investigation by mutagenesis.
In the case of the glutamates, each residue was changed to Ala, Gln, and Asp, whereas in the case of the arginines, each was changed to Ala, Gln, and Lys. The proteins were expressed and purified to homogeneity based on SDS-PAGE analysis using the WT protocols (supplemental Fig. S1). In the case of the ␤2 mutants, the diferric-Y 122 ⅐ was self-assembled to give a cofactor with a Y 122 ⅐ content similar to WT-␤2 (Table 1). All mutants were assayed for activity and a K d for each ␣/␤ interaction was determined ( Table 1). The E52X-␤2 (X ϭ Ala, Asp, or Gln) mutants have activity ϳ0.15% of WT-␤2, within the levels typically observed for endogenously copurifying WT-␤2. The K d measurements revealed that the Ala and Asp mutants are 5and 10-fold higher than WT, whereas Gln is similar to WT ( Fig.  2A). These studies suggest that Glu 52 plays an important role in catalysis.

N 3 CDP as a probe of E52X-␤2 (X ‫؍‬ Ala, Asp, or Gln)
Because RNR is essential, the issue of endogenous WT-RNR co-purifying with the mutants always hinders determination of a lower level of enzymatic activity. An alternative way to assess activity has been to use the mechanism-based inhibitor 2Ј-azido-2Ј-deoxycytidine diphosphate (N 3 CDP) (13,33). This NDP analog binds in the active site and is enzymatically con-verted to a nitrogen-centered nucleotide radical (N ⅐ ), that becomes covalently bound to a cysteine in the active site. The inactivation is stoichiometric with the WT-␤2, with complete loss of activity resulting from 1 Tyr 122 ⅐ /␤2 being converted to 0.5 eq of N ⅐ , leaving 0.5 eq of the Tyr 122 ⅐ remaining (34 -36). This unusual stoichiometry is associated with the half-sites reactivity of all class I RNRs. The N ⅐ has been extensively characterized by isotopic labeling and EPR methods. With mutant ␤2s the rate of formation of N ⅐ is often slow and the radical is quenched slowly with time; the kinetics often preclude N ⅐ detection, thus analysis of total radical loss as a function of time is monitored (13). The results of experiments in which E52X-␤2/␣2/N 3 CDP/TTP (X ϭ Ala, Asp, and Gln) were incubated and analyzed by EPR over 120 min are summarized in Fig. 3A (33). No N ⅐ is observed and the total Tyr 122 ⅐ varies no more than 10% over the 2-h time period. With WT-␤2, 0.5 eq of N ⅐ is formed within 30 s. Thus, no activity of E52X mutants is apparent by this method either.
A third method to assess RNR activity is to place E52X into a different background: specifically one in which the Tyr 122 ⅐ is replaced with F 3 Y 122 ⅐ . The F 3 Y 122 ⅐ -␤2 mutant when incubated with ␣2/CDP/ATP has been studied in detail and shown to generate dCDP and the pathway Tyr 356 ⅐ (Fig. 1) at 25 s Ϫ1 in the first turnover and then reoxidize the putative F 3 Y 122 -O Ϫ to the F 3 Y 122 ⅐ in the rate-limiting step in the steady-state (20). This mutant is a hotter oxidant than Tyr 122 ⅐ and disrupts conformational gating of the RT process (20,32,37). The E52Q mutant in this background has 0.1% WT activity (Table 1), likely associated with endogenous levels of co-purifying WT-␤2. Thus all assays pointed to inactivity of E52Q-␤2.  (49) and K d was determined by the competitive binding method (12). All data are representative of at least two independent experiments. b Previously reported (12). c The wild-type NrdB that co-purifies with mutants may cause the low activity.  Efforts to determine the K d for subunit interactions with this double mutant gave data distinct from the single E52Q mutant (Fig. 2C) and the other mutants (Figs. 2A). The sharp break suggests a "stoichiometric" titration. Reanalysis of these data in which activity is monitored with increasing concentrations of E52Q/F 3 Y 122 ⅐ -␤2 reveal that for 0.1 M ␣2␤2 complex, 0.28 M of the double mutant was required for complete inactivation (Fig. 2D). Given that the mutant protein used in this experiment has 0.7 F 3 Y 122 ⅐ /␤2 with the radical equally distributed between the two ␤ monomers and assuming that the diferric-cluster without radical binds much more weakly, then one would predict the requirement for 0.29 M mutant, very similar to the experimental observation.

CDP/ATP, GDP/TTP, and N 3 CDP/TTP to probe E52Q/F 3 Y 122 ⅐ -␤2 activity by EPR methods
Although no activity of E52Q-␤2 or E52Q/F 3 Y 122 ⅐ -␤2 was observed under steady-state conditions, additional experiments were performed on E52Q/F 3 Y 122 ⅐ -␤2 to determine whether chemistry could be observed in the first turnover. As noted above, addition of CDP/ATP/␣2 to the single mutant, F 3 Y 122 ⅐ -␤2, results in formation of Tyr 356 ⅐ and a burst of dCDP (0.5 eq/F 3 Y 122 ⅐ ). The double mutant, E52Q/F 3 Y 122 ⅐ -␤2 was incubated with CDP/ATP and analyzed by EPR spectroscopy for production of the Tyr 356 ⅐ . The results are shown in Fig. 4A and are summarized in Table 2. The data reveal that only 4% of the total radical is lost within 1 min and that it increases to 30% by 5 min. Also within the 1-min time frame, 0.50 eq of Tyr 356 ⅐ is formed. The rate of loss of the total radical is substantially reduced when CDP is omitted. When ATP is omitted, however, the total radical is reduced to 50% by 5 min and the amount of Tyr 356 ⅐ is increased to 40% of the total radical by 1 min and remains unchanged at 5 min. Thus CDP is the predominant driver of Tyr 356 ⅐ formation and the effector (ATP) appears to stabilize the F 3 Y 122 ⅐ radical in ␤2 when no substrate is present for reduction.
An identical set of experiments carried out with purine substrates and effectors have the same phenotypes. The results are summarized in Table 2. With GDP/TTP, by 5 min 30% of the total radical is lost, whereas 0.5 eq of Tyr 356 ⅐ is formed within 2 min. The effector TTP stabilizes total radical and limits Tyr 356 ⅐ formation, whereas GDP is the predominant driver of Tyr 356 ⅐ .
What is most amazing about these results is that under steady-state conditions where neither E52Q nor E52Q/F 3 Y 122 ⅐ -␤2 make dCDP, E52Q/F 3 Y 122 ⅐ -␤2 can initiate RT subsequent to S/e binding.  ⅐ -␤2, WT-␣2, CDP, and ATP as a reference (20). C, subtraction of F 3 Y 122 ⅐ (red) from the composite spectrum at 10 min (black) from the reaction of E52Q/F 3 Y 122 ⅐ -␤2, WT-␣2, N 3 CDP, and TTP reveals the spectrum in blue. D, spectrum of N ⅐ observed in the reaction of WT-␤2, WT-␣2, N 3 CDP, and TTP as a reference (33). Table 2 Reaction of E52Q/F 3 Y 122 ⅐ -␤2 a and WT-␣2 a with either ATP/CDP or TTP/ GDP or TTP/N 3 CDP analyzed by EPR spectroscopy a The concentration, 15 to 50 M of 1:1 E52Q/F 3 Y 122 ⅐ -␤2 and WT-␣2. b The spectrum after subtraction was similar to background. c ND, not determined.
As noted above, a second way to look for activity, uses N 3 CDP or N 3 CDP/TTP. The results of this set of experiments are shown in Fig. 4C and summarized in Table 2. In contrast to the results with the single mutant (E52Q), N ⅐ is formed and accounts for 49% (N 3 CDP/TTP) versus 43% (N 3 CDP) of the total radical at 10 min (compare Fig. 4, C with D, an authentic standard for N ⅐ ). Thus these data also support the activity of the double mutant, E52Q/F 3 Y 122 ⅐ -␤2, at least on the first turnover.

E52Q/F 3 Y 122 ⅐ -␤2 with pre-reduced ␣2, CDP, and ATP can produce dCDP
The above observation that the double mutant, E52Q/ F 3 Y 122 ⅐ -␤2, is capable of RT to the ␣2 catalytic site suggests that this protein may be able to make dCDP, even though no (or very low) activity is observed in the steady-state. To test for dCDP formation, an assay was carried out with a 1:1 ratio of subunits at 20 M in the presence of CDP alone (blue), CDP/ATP (red), and CDP/ATP with reductant TR/TRR/NADPH (green) and the reaction was monitored as a function of time (Fig. 5). The amount of the Tyr 356 ⅐ (0.5 eq) observed (Table 2) is likely formed during reverse RT and suggested that 0.5 eq of dCDP would be generated. The results shown in Fig. 5 suggest that this is the case. There is a burst of dCDP formation and it is independent of the presence of reductant. The size of the burst in all three experiments is similar to the amount of Tyr 356 ⅐ formed, consistent with half-sites reactivity and one turnover. In all experiments, the burst phase is followed by a slow phase that occurs from 0.2 to 0.6% (1.6, 3.4, and 4.4 nmol/min/mg in Fig. 5, blue, red, and green, respectively) of that observed with the single mutant, F 3 Y 122 -␤2 (686 nmol/min/mg). The rate is fastest with TR/TRR/NADPH/CDP/ATP Ͼ CDP/ATP Ͼ CDP. A number of explanations are possible for this slow phase observed in all experiments. In the absence of reductant (red and blue, Fig. 5) the slow phase could be associated with endogenous ␤2 acting catalytically, with very slow completion of the catalytic cycle in which Tyr 356 ⅐ must reoxidize the F 3 Y-O Ϫ or with slow release of cytosine catalyzed by the oxidized form of RNR. This issue remains unresolved. However, the interesting result is that E52Q/F 3 Y 122 ⅐ -␤2 is able to carry out one turnover! Thus, although the steady-state assays do not reveal significant activity (0.1% WT, Table 1), the double mutant is capable of the radical-based reactions that result in dCDP formation.

Interaction of His 6 -␣2 and E52Q/F 3 Y 122 ⅐ -␤2 using pulldown assays and SDS-PAGE analysis
Our previous studies showed that incubation of His 6 -NH 2 Y 730 -␣2 with ␤2, CDP, and ATP resulted in formation of NH 2 Y 730 ⅐ concomitant with Tyr 122 ⅐ loss. Rapid purification of His 6 -NH 2 Y 730 -␣2 from this mixture using a Ni-NTA affinity resin by centrifugation followed by SDS-PAGE analysis showed that ␣ and ␤ co-purified (24).
Centrifugation analysis monitoring supernatants from time 0 to 120 min incubation prior to workup revealed that when no CDP was present (time 0), no E52Q/F 3 Y 122 ⅐ -␤2 was pulled down, but within 1 min of its addition, the pulldown was maximized and remained unchanged (Fig. 6A, right). The majority of the pulldown experiments were carried out using a column gravity workup (Fig. 6B), as it typically gave higher recoveries of His 6 -␣2 (Ͼ80%). A variety of experiments were carried out in which the S (CDP or GDP), e (ATP or TTP), or S/e pairs and the incubation times, 5 or 30 min, were varied. In addition, controls with ␤2, E52Q/Y 122 ⅐ -␤2, and F 3 Y 122 ⅐ -␤2, or E52Q/Y 122 ⅐ -␤2 without S/e were also examined. The results summarized in Table 3 reveal that with S alone or S/e that a ␤2/␣2 ratio of 0.5-0.8 was observed, where with e alone, the ratio was lower at 5 min, but increased by 30 min (experiments 8 and 12). The data together suggest that the appropriate S/e pair form

Importance of glutamate 52 in ␤ of class Ia RNR
"tight" complexes rapidly and that tight complex remains at 30 min. These conclusions are supported by the controls (Table 3, 1-4) that all have low ␤2/␣2 ratios, 0.0 -0.2, in the pulldowns. These studies suggest the F 3 Y 122 ⅐ , a conformational uncoupler that generates the Tyr 356 pathway radical in combination with the E52Q mutation are important for successful ␣2␤2 complex formation.

Characterization of the reaction mixture by SEC and negative stain EM
Two additional types of experiments were carried out to support an ␣2␤2 complex structure and the tightness of the complex. In one set of experiments the reaction of E52Q/F 3 Y 122 ⅐ -␤2 was incubated with 0.5 eq of ␣2 (1:2, ␣2:␤2 subunit ratio), GDP, and TTP and loaded on a Superdex 200 SEC column and then eluted with assay buffer containing 50 M GDP and 10 M TTP. The results shown in Fig. 7A reveal a peak eluting at 12.1 ml and a broad peak at 13.7 ml. Comparison with molecular weight standards in Fig. 7B suggests that the former is ␣2␤2 and the latter is ␤2 and the ratio is 1:1 based on a comparison of the relative peak areas as expected from experimental design (Fig.  7, red). When the FPLC experiment was carried out in the absence of nucleotides in the elution buffer, peaks were observed at very similar elution volumes (Fig. 7, black), but the ratio of the peak intensities suggest only ϳ40% ␣2␤2 complexation. In a control with F 3 Y 122 ⅐ -␤2/␣2/GDP/TTP, no ␣2␤2 complex was observed (Fig. 7A, blue). Control experiments with E52Q/Y 122 ⅐ -␤2 in place of E52Q/F 3 Y 122 ⅐ -␤2 showed ␣2␤2 complex formation with GDP/TTP in the elution buffer, whereas no ␣2␤2 was observed without GDP/TTP (not shown).
In these experiments, the peaks corresponding to ␣2␤2 at 12 ml eluted 25 min after reaction initiation with GDP and TTP. Thus although the pulldown experiments allow isolation of ␣2␤2 with very high recovery and no GDP/TTP in the elution buffer, the SEC data tell us that on the 30-min time scale of the SEC analysis, the two subunits come apart in the absence of nucleotides during chromatography.
In a second set of experiments, ␣2␤2 complex formation was examined by negative stain EM. Our previous studies on the reaction of NH 2 Y 730 -␣2, ␤2, CDP, and ATP reported our first efforts to look for the "active" ␣2␤2 complex by this method (24). The resulting low resolution (ϳ32 Å) model revealed a subunit arrangement that was consistent with the ␣2␤2 docking model (Fig. 1). Interestingly, when WT-␣2 and WT-␤2 were mixed and observed on an EM grid with negative stain, almost all observed particles were of free ␣2 and almost no ␣2␤2 complex was observed. Free ␤2 is too small (87 kDa) to be visualized. NH 2 Y 730 -␣2 with WT-␤2 gave rise to ϳ70% ␣2␤2 particles (24).
Here, negative stain EM experiments with WT-␣2 and E52Q/F 3 Y 122 ⅐ -␤2 with GDP/TTP were carried out under similar conditions to the SEC (Fig. 7) and pulldown (Table 3) experiments. What is immediately striking is the large number of ␣2␤2 complexes that are present (Fig. 8), estimated to be 90%. The ratio of 1:2 for ␣2:␤2 was chosen to maximize the chemistry (Tyr 356 ⅐ formation) as typically there are ϳ0.8 E52Q/ F 3 Y 122 ⅐ -␤2. Taken together, the pulldown studies, EM, and SEC analysis reveals ␣2␤2 complexes that are supported by biochemical analysis that shows active RT and dCDP formation. The SEC data reveal that further work, such as our stopped flow fluorescence studies on NH 2 Y 730 -␣2, will be informative in determining a quantitative assessment of the subunit affinity in the complex observed. ⅐ -␤2 with ␣2, GDP, and TTP in the presence (red) or absence (black) of nucleotides in the elution buffer and a control experiment with F 3 Y 122 ⅐ -␤2, ␣2, GDP, and TTP in the presence of GDP/ TTP in eluent (blue). A, the peak eluting at 12.1 min has a molecular weight consistent with ␣2␤2, whereas the broad peak at 13.7 min is likely uncomplexed ␤ and ␣. The experiment was carried out under the same conditions as the negative stain EM images. A 1:2 ratio of ␣2:␤2 was used to maximize complex formation. B, molecular mass standards are ferritin (440 kDa), aldolase (158 kDa), conalbumin (75 kDa), and ovalbumin (44 kDa).

Activity and subunit binding affinity of additional ␣ and ␤ mutants (Table 4)
In addition to Glu 52 , our search for charged interface residues suggested that Arg 329 and Arg 639 in ␣ were also of interest. Arg 329 is located in loop 3 of ␣ and is adjacent to a second Arg at 323 that is not conserved (see supplemental Fig. S2). Mutants of Arg 329 (Ala, Lys, and Gln) were made and assayed for activity and binding to WT-␤2. The results of mutation of Arg 329 to Ala, Lys, and Gln and a control of Arg 323 to Lys are shown in Table 4. R329A and R329Q have no detectable activity, whereas R329K has 0.12% activity of WT-␣2. Binding studies (12) revealed that all three mutants exhibit 10 -20-fold weaker binding than WT, similar to the phenotypes of the E52X mutants with the exception of Gln (Fig. 2, A and B). Given the weak K d values, the mutants were assayed at higher protein concentration and still found to have no activity (Table 4). In the case of R329K, an additional experiment was carried out with N 3 CDP to look for N ⅐ formation and total radical loss. These results (Fig. 3B) also indicate that this mutant is inactive and hence important. The control, a lysine mutant of Arg 323 has 34% the activity of the WT-␣2 and was thus not considered further.
The R329X-␣ mutants were also studied with F 3 Y 122 ⅐ -␤2. Activity of 1 to 3% WT was observed with the Lys mutant having the highest level (Table 4). These studies also suggest that Arg 329 plays an important role in the RNR catalysis. Finally, studies with an R639Q-␣ mutant revealed that it is inactive, whereas mutations of the non-conserved Arg 735 , also proposed to be at the interface, results in active enzyme.

Discussion
In the past decade using technology to site-specifically incorporate UAA coupled to time-resolved kinetic measurements to study the consequences of the incorporation, much has been learned about the long distance RT process required to initiate nucleotide reduction in RNR (3,26,38). The UAAs have been one of the crucial perturbants to allow uncoupling of the ratelimiting conformational gating that masks the RT and the nucleotide reduction chemistry in the WT enzyme (20).
Binding of the appropriate S and e pairs to ␣2 followed by binding ␤2 has long been known to trigger the essential conformational change(s) that occurs over the 37 Å (C-site) or 39 Å (S-site) in ␣2 to the RT initiation site in ␤2 (Fe 1 to S of Cys 439 or 2-O of TTP, respectively) (19,32). The binding of CDP/ATP or CDP (or GDP/TTP or GDP) changes the loop 2 structure in ␣2 and also induces a closure of the barrel structure around the catalytic site on ␣2 (39). These changes must be transmitted across the ␣/␤ subunit interface, likely through a conserved network of residues to position the water bound to the Fe 1 in the diferric cluster, so that the proton can be efficiently delivered to Tyr 122 ⅐ concomitant with its reduction (Fig. 1).
Our recent studies on the conserved Glu 350 in ␤2, suggested that it likely plays a very important role in conformational gating (3,18). We thus decided to investigate the possible role of other conserved interface residues including Glu 52 -␤2, Arg 329 -␣2, and Arg 639 -␣2.
The results with the E52Q in the WT and F 3 Y 122 ⅐ backgrounds are most striking. In the WT background it is unable to make dNDPs and is inactive in the formation of the N ⅐ from N 3 CDP. E52Q-␤2, in contrast to the Glu and Asp mutants, binds similarly to ␣2 as WT-␤2 (Table 1, Fig. 2A) and for E52Q/ F 3 Y 122 ⅐ -␤2 the binding to ␣2 is stoichiometric (Fig. 2, C and D). Additionally, although the E52Q/F 3 Y 122 ⅐ -␤2 is inactive in the steady-state assay, it is able to make N ⅐ from N 3 CDP, Tyr 356 ⅐ in the presence of CDP/ATP, and catalyze 1/2 turnover (one CDP/two F 3 Y 122 ⅐ ), consistent with the half-site reactivity of RNR (20,23). It is likely that the reoxidation of F 3 Y 122 -O Ϫ to the F 3 Y 122 ⅐ by Tyr 356 ⅐ is too slow to compete with loss of the total radical (Table 2) (18,20,37), potentially explaining the lack of activity under steady-state conditions. This reoxidation is also slow for F 3 Y 122 -O Ϫ in the WT background, but the E52Q mutation appears to result in an even slower process.
To investigate ␣/␤ binding, we used several pulldown approaches. The experiment with His 6 -␣2, S/e (where S is CDP or GDP and e is ATP or TTP), and E52Q/F 3 Y 122 ⅐ -␤2 allowed isolation of a complex by Ni-NTA affinity chromatography with a ␤2/␣2 subunit ratio of ϳ0.6. In contrast, the ratio of 0.01 was observed with the WT control after a 5-min incubation (Table 3).
Interestingly, the double mutant complex has a longer lifetime than the pathway (Tyr 356 ⅐ ) radical in the pulldown assays. The total amount of radical (F 3 Y 122 ⅐ and Tyr 356 ⅐ ) decreases 30 to 50% over 5 min (Table 2), yet the complex can be isolated over 2 h (Fig. 6A, right, and time course data not shown with the other experiments in Table 3). Thus, the conformation of the ␣2␤2 complex that allowed its isolation appears to have a "kinetic" memory, that is, it remains in an altered conformation Table 4 Specific activities for mutant-␣2s with 5-fold WT-␤2 or 10-fold F 3 Y 122 ⅐ -␤2 determined by the radioactive assay (49) and K d for mutant-␣2/WT-␤2 interaction determined by the competitive inhibition assay (12) All data are representative of at least two independent experiments.  . (12). b The counts were the same as the background control. c ND, not determined.

Importance of glutamate 52 in ␤ of class Ia RNR
after much of the pathway radical has decayed. This observation of a kinetic memory is strikingly similar to our recent studies with ␣ on the human RNR. This subunit forms a hexameric structure, ␣6, in the presence of dATP or the phosphorylated drugs clofarabine di-or triphosphate (ClFDP or ClFTP) (40 -42). When dATP dissociates from ␣6, the hexamer returns to a monomeric state. However, when ClFDP or ClFTP dissociate, the hexameric structure remains. The molecular basis for the continued tight binding of ␣2␤2 in the case of the E. coli RNR double mutant and ␣6 in the hRNR remain unknown. However, it is intriguing in the case of the E. coli RNR that a conservative chemical substitution Gln for Glu in the F 3 Y 122 ⅐ -␤2 has such a dramatic effect on ␣/␤ interactions in pulldown assays. From the many ␤2 structures available, we know that Glu 52 located on the surface of ␤ is conformationally flexible with "out," "in," and "intermediate" conformations (supplemental Fig. S2, B and C). Its "in" conformation connects through waters to a conserved residue, Arg 236 , within ␤. Arg 236 has connectivity to Trp 48 that in turn connects to Asp 237 , which connects to His 118 , a ligand to Fe 1 of the cofactor (supplemental Fig. S2C). It is the water on Fe 1 that is proposed to deliver the proton to Tyr 122 ⅐ upon Tyr 122 ⅐ reduction (supplemental Fig. S2C) (11,19,43). Also shown in supplemental Fig. S2 is the location of the "out" conformation of Glu 52 relative to the conserved Arg 329 in loop 3 of ␣ in the ␣2␤2 docking model. Supporting the importance of Arg 329 , mutants (Gln, Lys, and Ala) show weak binding to ␤2, with K d values elevated 10-fold relative to WT, similar to the results with Glu 350 and Glu 52 mutants. The inactivity of Glu 52 and Arg 329 mutants might result from their altered conformations in this region of ␣2. The studies with E52Q/ F 3 Y 122 ⅐ -␤2 and the requirement for S/e suggest its importance in conformational triggering of RT across ␣/␤. The unexpected observation of the high percentage of the ␣2␤2 complex formed in the double mutant may provide the opportunity to gain insight into the structure of this complex based on our negative stain EM images (Fig. 8).
Finally, the least well studied mutant, Arg 639 -␣ has very low activity and has weakened binding to ␤. Recent structures from the Drennan lab (39) show that in the presence of the correct S/e pairs, loop 2 (yellow, supplemental Fig. S3A) becomes ordered, the barrel clamps around the catalytic site, and the ␤-hairpin (supplemental Fig. S3B, blue to orange) moves to potentially protect the active site. Arg 639 , which is adjacent to this hairpin may play a role in stabilizing the differential hairpin conformations. Interestingly, this ␤-hairpin is conserved in the class II RNRs and is observed to move when the adenosylcoblamin cofactor, the radical initiator, binds to initiate nucleotide reduction via formation of a thiyl radical (44).

Conclusions
The reversible long distance RT between ␣ and ␤ continues to be a fascinating feature of the class I RNRs. RT is gated subsequent to binding the appropriate S and e pairs on the ␣ subunit, requiring communication across the subunits over a distance of 35 to 40 Å. The transient nature of the ␣ and ␤ interactions in the E. coli RNR, the flexibility of its ␣ and ␤ tails both essential in catalysis, the complexity and number of nucleotide-binding sites, have all made an understanding of the molecular mechanism of conformational gating and a structure of an active RNR elusive. Here we have identified conserved residues likely to control conformational gating at the ␣/␤ interface. The most intriguing results are that the double mutant of E52Q/F 3 Y 122 ⅐ -␤2 when incubated with ␣2, S, and e, potentially forms the "tightest" complex thus far reported based on pulldown assays, SEC, and negative stain EM studies. The conservative mutation of Glu 52 to Asp, on the other hand, weakens subunit affinity compared with WT. Clearly the design of the subunit interface is intricate, providing the exquisite control that is needed for the RT chemistry mediated by S/e in this essential enzyme.

RNR activity assays
The

Time-dependent inactivation of RNR mutants in the presence of N 3 CDP
A 250-l reaction mixture contained: protein (30 M WT-␣2 with 30 M E52X-␤2 (X ϭ Ala, Asp, or Gln) or 30 M R329X-␣2 (X ϭ Ala, Lys, or Gln) with 30 M WT-␤2], 0.2 mM TTP, 50 mM Hepes (pH 7.6), 1 mM EDTA, 15 mM MgSO 4 and was incubated at 25°C for 1 min. The time 0 sample was frozen in liquid nitrogen and the EPR spectrum was recorded. The sample was then thawed and the reaction started by addition of 0.25 mM N 3 CDP. The control had no N 3 CDP. Each sample was warmed to 25°C and used for a complete time course study by repeated freezethaw cycles (50). The amount of radicals were quantitated as previously described (28).

K d measurements for the interaction between ␣2 and ␤2 mutants
The interaction between E52X-␤2 (X ϭ Ala, Asp, or Gln) for ␣2 and R329X-␣2 (X ϭ Ala, Lys, Gln, or Glu) for ␤2 were determined in assay buffer at 25°C by the competitive inhibition assay (12). Efforts to determine the affinity of E52Q/F 3 Y 122 ⅐ -␤2 for ␣2 were carried out by the same procedure. However, the binding curves could only be fit under the assumption that the subpopulation of E52Q/F 3 Y 122 -␤2 lacking F 3 Y 122 ⅐ does not competitively inhibit in the concentration range of the experiment (Fig.  2C). An upper limit of the K d was estimated from this model (  (49).

Reaction of E52Q/F 3 Y 122 ⅐ -␤2 with WT-␣2 monitored by EPR spectroscopy
In a final volume of 250 l the reaction mixture contained WT-␣2 (15 to 50 M), with 1 eq of ␤2, substrate (CDP (1 mM) or GDP (1 mM) or N 3 CDP (0.25 mM)), Ϯ effector (ATP (3.0 mM) or TTP (0.2 mM)) in assay buffer. Samples were incubated for a specified time in a circulating water bath at 25°C and quenched for EPR analysis in liquid nitrogen. EPR spectra were recorded at 77 K in the Department of Chemistry Instrumentation Facility on a Bruker ESP-300 X-band spectrometer equipped with a quartz finger Dewar filled with liquid nitrogen. Typical EPR parameters were as follows: microwave frequency ϭ 9.45 GHz, power ϭ 32 W, modulation amplitude ϭ 1.5 G, modulation frequency ϭ 100 kHz, time constant ϭ 40.96 ms, scan time ϭ 41.9 s. Analysis of the resulting spectra was carried out using WinEPR (Bruker) and an in-house written program in Matlab R2016a (50). EPR spin quantitation was carried out using Cu II as standard.

Pulldown assays
A final volume of 100 l contained untagged-␤2s (10 M), His 6 -WT-␣2 (10 M), ATP (3 mM), or TTP (0.2 mM) in assay buffer at 25°C. CDP (1 mM) or GDP (1 mM) or alternatively mutant ␤2 was added to initiate the reaction. The reaction mixture was incubated for 1 to 120 min at 25°C and then combined with a nickel-nitrilotriacetic acid resin (ϳ60 or 300 l, from Qiagen) suspended in the EDTA-free assay buffer and rotated by hand at room temperature for 1 min. The sample was then centrifuged (30 s, 3,000 ϫ g, 4°C) and the supernatant was removed. Alternatively, the NTA resin (300 l) was placed in a small column and eluted by gravity. In the former case, the resin "pellet" was rapidly resuspended in 600 l of wash buffer (EDTA-free assay buffer with 300 mM NaCl and 15 mM imidazole (pH 7.6)) and centrifuged (30 s, 3,000 ϫ g, 4°C). This wash step was repeated a second time. Resin-bound protein was then eluted by resuspending it in elution buffer (100 l, EDTA-free assay buffer with 250 mM imidazole (pH 7.6)), followed by centrifugation (30 s, 3,000 ϫ g, 4°C). The procedure (flow through, washes (W1 and W2), and elution (E)) took 5 min. The recovery of ␣ is typically 40 to 50%.
In the latter case, gravity elution, the procedure (loading, washes, and elution) is the same except that the procedure takes 2 to 3 min and the recovery of ␣ is typically ϳ90%. The contents of each fraction were assessed by SDS-PAGE (10%) and compared with the fractions obtained in a control experiment with standards made from stock solutions: 1 M His 6 -WT-␣2 and 1 M WT-␤2.

Negative stain EM on ␣2 with E52Q/F 3 Y 122 ⅐ -␤2
A reaction mixture was prepared with 5 M ␣2, 10 M E52Q/ F 3 Y 122 ⅐ -␤2, 1 mM GDP, and 0.2 mM TTP in assay buffer (50 mM Hepes, pH 7.6, 15 mM MgSO 4 , and 1 mM EDTA) where ␤2 was added last to initiate the reaction. The mixture was incubated 3 min at 25°C and then diluted 130-fold in assay buffer containing 1 mM GDP and 0.2 mM TTP giving final protein concentrations of 40 nM ␣2 and 80 nM E52Q/F 3 Y 122 ⅐ -␤2. The solution was applied to a 300-mesh continuous carbon grid (EMS) and stained three times with a 1% uranyl acetate solution. The total Importance of glutamate 52 in ␤ of class Ia RNR time between reaction initiation and application onto the grid was ϳ15 min.

Data collection
All images were collected at the W. M. Keck Institute for Cellular Visualization at Brandeis University. The grids were imaged at 200 kV on a Tecnai F20 electron microscope (FEI) equipped with an UltraScan 4000 CCD camera (Gatan) using SerialEM operated in manual low-dose mode at a magnification of 62,000 with a pixel size of 1.79 Å at the specimen level.