Role of the C terminus of the ribonucleotide reductase large subunit in enzyme regeneration and its inhibition by Sml1

Abstract

Ribonucleotide reductase maintains cellular deoxyribonucleotide pools and is thus tightly regulated during the cell cycle to ensure high fidelity in DNA replication. The Sml1 protein inhibits ribonucleotide reductase activity by binding to the R1 subunit. At the completion of each turnover cycle, the active site of R1 becomes oxidized and subsequently regenerated by a cysteine pair (CX₂C) at its C-terminal domain (R1-CTD). Here we show that R1-CTD acts in trans to reduce the active site of its neighboring monomer. Both Sml1 and R1-CTD interact with the N-terminal domain of R1 (R1-NTD), which involves a conserved two-residue sequence motif in the R1-NTD. Mutations at these two positions enhancing the Sml1–R1 interaction cause SML1-dependent lethality. These results point to a model whereby Sml1 competes with R1-CTD for association with R1-NTD to hinder the accessibility of the CX₂C motif to the active site for R1 regeneration.

The maintenance of adequate and balanced dNTP levels is critical for faithful DNA replication and repair and for the survival of all organisms (1–5). A major target of dNTP pool regulation is ribonucleotide reductase (RNR) that catalyzes the essential step of converting ribonucleoside diphosphates to the corresponding deoxy forms in dNTP biosynthesis (6). Class I RNRs are conserved from eubacteria to eukaryotes and comprise two subunits: R1 and R2. R1 (α₂ in Escherichia coli and α₂, α₄, and α₆ in eukaryotes; ref. 7) contains the active site as well as binding sites for allosteric effectors; R2 (β₂) houses a diferric-tyrosyl radical required for catalysis (8). These subunits can be regulated by allostery (1), transcription (9), subcellular compartmentalization (10–13), and protein inhibitor interaction (14, 15).

The 104-residue Saccharomyces cerevisiae Sml1 protein was originally identified as an RNR inhibitor based on the finding that loss of SML1 function suppresses the lethality of cells lacking the checkpoint kinases Mec1 or Rad53 by increasing cellular dNTP levels (15). Sml1 is phosphorylated and degraded during S phase and after DNA damage in a checkpoint-dependent manner to relieve RNR inhibition (16). The inhibition of R1 by Sml1 depends on Sml1–R1 association because mutations in SML1 disrupting its R1-binding ability abolish the inhibition (17). Crystallographic studies of the R1s from E. coli and S. cerevisiae reveal three domains in the protein: the N-terminal helical domain, the 10-stranded α/β-barrel domain, and the C-terminal domain of less-defined structure (18, 19). The active site is located in the center of the protein between the N-terminal and the barrel domains, wherein a redox-active cysteine pair (Cys-225/Cys-462 of the E. coli R1 and Cys-218/Cys-443 of the yeast R1) converts from a free dithiol form in the reduced R1 (active form) to a disulfide-bonded form in the oxidized R1 (inactive form) after each reduction cycle (20). This disulfide bond is reduced to regenerate an active R1 for the subsequent catalytic cycles (21, 22). The physiological reductants for R1 regeneration are thioredoxin and glutaredoxin (23, 24), although these two proteins cannot interact directly with the R1 active site (22, 25, 26). In vitro studies suggest that a conserved cysteine pair at the R1 C-terminal end (designated as the CX₄C motif in the bacterial R1s or CX₂C in the eukaryotic R1s) may act as an intermediate in a two-step disulfide exchange reaction, with the active-site cysteine pair and thioredoxin/glutaredoxin to achieve R1 regeneration (22, 25, 26). However, this hypothesis has not been tested in vivo. It is also not known whether the C terminus of each R1 monomer interacts with the active site of its own (in cis) or that of its neighboring monomer (in trans) through a cross-talk mechanism to effect reduction (26, 27).

There are two R1 isoforms in S. cerevisiae: Rnr1 (α) and Rnr3 (α′) (28). Only Rnr1 is essential for survival, and the level of Rnr3 is normally low but highly inducible after DNA damage (28). The N-terminal ∼780-residue domain (R1-NTD) is highly conserved between Rnr1 and Rnr3 as well as among most of the eukaryotic R1s. Interestingly, the yeast R1 proteins contain an extended C-terminal domain (R1-CTD) comprising a ∼100-residue C-terminal insert (CI) region and the CX₂C motif at the C-terminal end (Fig. 1A). In this work, we provide evidence that the CX₂C motif is essential for RNR function in vivo. Interallelic complementation analyses indicate that the R1-CTD of one R1 monomer acts in trans to regenerate the active site of its neighboring monomer. Moreover, we show that Sml1 and R1-CTD compete for interaction with R1-NTD via a highly conserved two-residue motif in the R1-NTD. Substitutions at the two positions strengthening the interaction between R1-NTD and Sml1 can result in SML1-dependent cell lethality. Taken together, our results point to a competition model whereby Sml1 inhibits R1 activity by impeding the access of the C-terminal CX₂C motif to the R1 active site necessary for regeneration.

Fig. 1.

Interaction between R1-NTD and R1-CTD. (A) Schematic alignment of the S. cerevisiae Rnr1 and Rnr3 proteins with the E. coli R1. The thiyl radical-generating cysteine (Cys-439 in E. coli and Cys-428 in yeast) and the cysteine pair at the C-terminal end are shown. Both Rnr1 and Rnr3 have a CI region. R1-CTD refers to the entire C-terminal region including the CI and the CX₂C motif. (B and C) Interaction between R1-CTD and R1-NTD in a yeast two-hybrid system. Plasmids expressing the Gal4 activating domain (ACT) domain alone (Vec) and ACT fusion proteins with full-length Rnr1 (R1-FL) and the C-terminal 765–888 region of Rnr1 (R1-CTD) were cotransformed in pairs with plasmids expressing the Gal4 DNA-binding domain (DBD) alone (Vec) and DBD fusion proteins with the full-length (R1-FL) and the N-terminal 778- and 768-residue regions of R1. Growth of cotransformants on SC-TrpLeu and SC-TrpLeuAde plates is shown in B. Activities of the LacZ reporter were measured in Miller units in C.

Results

Interaction Between R1-CTD and R1-NTD.

In all known crystals structures of R1, the C-terminal regions are thermally labile and structurally undefined (18, 19, 29, 30). Thus, the location(s) of the C terminus in the tertiary structure of R1 is not known. To probe domain–domain interaction within R1, we exploited the yeast two-hybrid system by fusing the Gal4 DBD to the full-length (R1-FL, 888 residues) and N-terminal 778 residues of R1 (i.e., R1-NTD), and making Gal4 ACT fusions to R1-FL and R1-CTD (residues 765–888). The ACT-R1 (765–888) fusion was used because it produces a stable expression level of fusion protein (ACT fusions with residues 762–888 and 778–888 of R1 did not produce stable proteins). We then tested the interactions between each pair of DBD and ACT fusion constructs in the PJ69–4a strain by scoring the Ade⁺ phenotype as an indicator of a positive interaction, followed by LacZ activity assays to probe the strength of the domain–domain interactions (31).

The active form of R1 is believed to be a dimer or oligomer of dimers (6, 7). As expected, we observed positive interaction between R1-FL itself (Fig. 1B). The N-terminal 778-residue R1-NTD also interacts with R1-FL, noticeably with higher affinity (Fig. 1C). Unexpectedly, ACT-R1CTD did not interact with DBD-R1FL. We reasoned that if the ACT-R1CTD interacted with a site within the R1, this site might be occupied by the R1-CTD present in a preformed R1 dimer, to obscure the detection of protein–protein interaction in trans. Consistent with this notion, we detected a positive interaction between ACT-R1CTD and DBD-R1NTD, wherein the R1-CTD is deleted in cis (Fig. 1 B and C). The observed interaction is unlikely to be mediated by the 14-aa region (residues 765–778 of R1) shared by the DBD-R1NTD and ACT-R1CTD fusion proteins because a DBD-R1 (1–768) fusion exhibited a positive interaction with R1-CTD similar to DBD-R1 (1–778) (Fig. 1 B and C). In summary, the data indicate that R1-CTD can associate with R1-NTD, consistent with the proposed role of R1-CTD in regeneration of the enzyme active site located at the center of R1-NTD (22).

The CX₂C Motif of Rnr1 Is Essential for Viability.

To determine the functional role(s) of the CX₂C and CI regions of the R1-CTD, we generated mutant alleles rnr1(CX₂C to SX₂S at 883–886), rnr1(Δ778–880), and rnr1(Δ778–888) and tested their ability to provide R1 activity in vivo. The Δ778–880 allele contained an in-frame deletion of the CI region, and the Δ778–888 allele deleted the entire R1-CTD, including the CI and CX₂C. The wild-type and mutant R1 proteins all contained an N-terminal (Myc)₃ epitope and were expressed from the RNR1 promoter on a centromeric plasmid (one or two copies per cell) (32). Yeast cells bearing the ^Myc3Rnr1 as the sole R1 were viable and exhibited growth rate and sensitivity similar to the potent RNR inhibitor hydroxyurea (Fig. 2B and data not shown). We then used a plasmid shuffle complementation assay (33) to examine the ability of these RNR1 alleles to support cell viability in an rnr1Δrnr3Δ double-mutant strain. Cells containing either the wild-type RNR1 or the rnr1(ΔCI) mutant allele are viable (Fig. 2B). By contrast, cells containing the rnr1(ΔCTD) or the rnr1(SX₂S) mutant alleles failed to form any colonies (Fig. 2B). The difference in the growth phenotype cannot be attributed to variation in protein levels (Fig. 2C). Although the ΔCI mutant was viable, it grows slowly. When released from an α-factor-mediated G₁ arrest, the mutant cells exhibited a prolonged S phase relative to cells bearing R1-FL (Fig. 3E). Taken together, our results provide in vivo evidence for an essential function of the CX₂C motif in R1, consistent with its proposed role in active-site regeneration based on biochemical studies of the E. coli RNR (22). Our results also suggest that the CI region, although dispensable for viability, is required for optimal function of R1.

Fig. 2.

The CX₂C motif of the S. cerevisiae Rnr1 is essential for viability. (A) Schematic drawings of the substitutions and deletions introduced into the S. cerevisiae Rnr1. (B) Plasmid shuffle complementation assay. Shown is the 5-FOA plate after 2 days of incubation at 30°C for transformants of the RNR1 shuffle strain MHY784 (rnr1Δ rnr3Δ URA3CENRNR1) containing the TRP1CEN vector or test plasmid expressing (Myc)₃-tagged wild-type and mutant Rnr1 proteins from the RNR1 promoter. (C) Comparison of protein abundance of different RNR1 alleles. The (Myc)₃-Rnr1 proteins were detected on a Western blot by using the 9E10 antibody (α-Myc). Glucose-6-phosphate 1-dehydrogenase (G6PDH by α-Zwf1) was also probed on the same blot as a loading control. (D) FACS analysis of cells harboring the full-length (FL) and the CI-deleted (ΔCI) alleles of RNR1 from asynchronous (Asy) or synchronized cultures after release from an α-factor-mediated G₁ arrest.

Fig. 3.

Interallelic complementation between the catalytically inactive rnr1(C428S) and the CX₂C-deficient rnr1(SX₂S) mutant alleles. (A) Growth on a 5-FOA plate after 2 days of incubation at 30°C for transformants of the RNR1 shuffle strain MHY784 containing the following RNR1 plasmids: wild-type (WT), rnr1(C428S), rnr1(CX₂C-to-SX₂S), and rnr1(CX₂C-to-SX₂S) in combination with rnr1(C428S). (B) Comparison of plating efficiency of the rnr1Δ rnr3Δ double mutant carrying the indicated RNR1 alleles on the rich medium YPD. Cells from a log phase culture of each strain were measured for density by using a hemocytometer and diluted so that ∼300 cells were plated on each plate. All plates were incubated at 30°C for 2 days before comparison of colony formation. (C) Comparison of protein levels between full-length and CI-deleted Rnr1 mutant proteins. Protein extracts were prepared from cells harboring the indicated rnr1 mutant alleles. The HA-tagged Rnr1(C428S) and Rnr1(C428S, ΔCI), and (Myc)₃-tagged Rnr1(SX₂S) and Rnr1(ΔCI, SX₂S) were detected on a Western blot by using anti-HA and anti-Myc antibodies, respectively. G6PDH (Zwf1) was probed on the same blot as a loading control.

Cross-Talk Between the C Terminus of One R1 Monomer and the Active Site of Its Neighboring Monomer.

We sought to determine whether the R1-CTD of one R1 monomer interacts with the active site of its own or that of its neighboring monomer through a cross-talk mechanism. The active-site Cys-439 of the E. coli R1 (counterpart of Cys-428 of the S. cerevisiae Rnr1) has been proposed to play an essential role in R1 function by forming a transient thiyl radical to initiate catalysis (26). Consistent with the finding that the C439S mutant of the E. coli R1 is catalytically inactive in vitro (26), we demonstrated that the rnr1(C428S) mutant allele failed to support mitotic growth of the rnr1Δrnr3Δ mutant (Fig. 3A). We therefore surmised that if cross-talk occurs between the R1 monomers, the intact CX₂C motif at the C terminus of the catalytically deficient Rnr1(C428S) mutant should provide the reducing activity toward the intact active site of the Rnr1(SX₂S) mutant lacking the CX₂C motif. To this end, we constructed two centromeric plasmids carrying the respective rnr1(C428S) and rnr1(SX₂S) mutant alleles under the control of the native RNR1 promoter and introduced them individually and in combination into the rnr1Δrnr3Δ mutant. In isolation, both rnr1(SX₂S) and rnr1(C428S) were lethal. Remarkably, cells carrying both the rnr1(C428S) and the rnr1(SX₂S) plasmids were viable and formed colonies with an efficiency similar to that of the wild-type control (Fig. 3A). This interallelic complementation points to a cross-talk interaction between the C terminus of one R1 monomer and the active site of its neighboring monomer during the active-site regeneration.

We then asked whether this cross-talk is unique to the S. cerevisiae R1, in part because the extended CI region might provide flexibility for the interchain protein–protein interaction. Accordingly, we constructed the rnr1(C428S, ΔCI) and rnr1(ΔCI, SX₂S) mutant alleles, each with a deletion of residues 778–880, and examined their abilities to reconstitute R1 function when combined with the full-length rnr1(C428S) and rnr1(SX₂S) alleles. Cells harboring the rnr1(C428S) + rnr1(ΔCI, SX₂S) pair showed similar growth phenotype in rich media compared with cells harboring the full-length rnr1(C428S) + rnr1(SX₂S) pair (Fig. 3B). Thus, deletion of the CI region in rnr1(ΔCI, SX₂S) does not influence the catalytic activity. Moreover, cells harboring the rnr1(C428S, ΔCI) mutant allele in combination with the rnr1(SX₂S) and the rnr1(ΔCI, SX₂S) are still viable (Fig. 3B), suggesting that the CI region is not absolutely required for the interchain cross-talk. We should nevertheless emphasize that the rnr1(C428S, ΔCI) allele exhibited slower growth phenotype and decreased plating efficiency when combined with rnr1(SX₂S) and rnr1(ΔCI, SX₂S), compared with rnr1(C428S) (Fig. 3B); little change in protein abundance was observed between the full-length and the CI-deletion proteins (Fig. 3C). Thus, the truncated CTD used in this work retains the ability to interact with the active site even though it could be less flexible to be fully functional.

Competitive Interaction of R1-NTD with R1-CTD and Sml1.

Given the role of R1-CTD in RNR active-site regeneration through a cross-talk mechanism, we wanted to address whether the R1 inhibitor Sml1 interferes with this cross-talk by testing the interaction of Sml1 with R1-FL and R1-NTD in a yeast two-hybrid system. Consistent with previous reports (15, 17), we observed a positive interaction between Sml1 and Rnr1-FL (Fig. 4A). Moreover, Sml1 interacted with R1-NTD. It is noteworthy that R1-NTD exhibited a stronger interaction with Sml1 relative to Rnr1-FL (Fig. 4A). The increased Sml1–R1 association in the absence of R1-CTD suggests that Sml1 and R1-CTD may compete for interaction with the conserved R1-NTD. In keeping with this notion, we have shown that overexpression of Sml1 from the strong TDH3 promoter (34) compromised the R1-NTD–R1-CTD interaction assayed by the yeast two-hybrid system (Fig. 4B).

Fig. 4.

Interaction of the R1-NTD with both Sml1 and R1-CTD. (A) Increased interaction of Sml1 with R1-NTD relative to R1-FL. Plasmids expressing the Gal4 ACT domain alone (pACT) and the ACT-Sml1 fusion protein (pACT-SML1) were cotransformed in pairs with plasmids expressing the Gal4 DBD fusion proteins with the full-length Rnr1 (Rnr1-FL) and the R1-NTD (residues 1–778). Activities of the LacZ reporter were measured in Miller units. (B) Overexpression of Sml1 compromises the interaction between R1-NTD and R1-CTD. Cells harboring pDBD-R1NTD and pACT, or pDBD-R1NTD and pACT-R1CTD were transformed with a plasmid expressing a TDH3 promoter-driven SML1 or the vector control. Activities of the LacZ reporter were measured in Miller units. (C) Substitutions of the WE 688–689 sequence motif in R1-NTD alter its interaction with both Sml1 and R1-CTD. (Upper) Sequence alignment of the region including the S. cerevisiae Rnr1 Tyr-688 and Glu-698 residues among the R1s of E. coli, human, and S. cerevisiae. (Lower) Two-hybrid interactions assayed with LacZ activities between the DBD-R1NTD fusion proteins containing the wild-type sequence (WE) or the WE-to-AD mutation, and ACT vector (Vec) and ACT fusion proteins with Sml1 or the R1-CTD (residues 765–888). (D) The WE-to-AD mutation at residues 688–689 of Rnr1 causes SML1-dependent lethality. Shown is the 5-FOA plate after 2.5 days of incubation at 30°C for transformants of the shuffle strains MHY784 (Upper) and MHY802 [rnr1Δ rnr3Δ sml1Δ URA3CENRNR1 (Lower)], all containing a TRP1CEN plasmid that carries the wild-type (WE) or the WE-to-AD mutant alleles of RNR1. (E) Overexpression of Rnr1 increases the Sml1 protein level. The endogenous Sml1 protein levels of wild-type cells harboring a plasmid expressing a TDH3 promoter-driven RNR1 or the vector control (vec) were probed on a Western blot. G6PDH (Zwf1) was probed on the same blot as a loading control. (F) The rnr1(WE-to-AD) mutant allele stabilizes the Sml1 protein after hydroxyurea (HU) treatment. Wild-type cells bearing an extra copy of the wild-type RNR1 or the rnr1(WE-to-AD) mutant allele, both under the control of the RNR1 promoter, were treated with 150 mM hydroxyurea, and the endogenous Sml1 levels were probed on a Western blot at the indicated time points after the addition of hydroxyurea. G6PDH (Zwf1) was probed on the same blot as a loading control.

The previously identified point mutation rnr1(W688G) partially restores the Sml1–R1 association disrupted by various point mutations in SML1 (35). Alignment of various R1 protein sequences revealed that Trp-688 and its neighboring residue Glu-689 are highly conserved from bacteria to human in the last ∼100 residues of the R1-NTD (Fig. 4C). These two residues are located at a small α-helical region present on the outside surface in the E. coli and yeast R1 crystal structures (18, 19). We posited that if this region is involved in interaction of R1-NTD with Sml1 and R1-CTD, alterations at these two positions might influence both of the protein–protein interactions. To test this notion, we created mutations in Trp-688 and Glu-689 and tested their effect in the R1-NTD interaction with Sml1 and α-CTD. We have identified one mutant allele, a WE-to-AD substitution, which enhanced the interaction of R1-NTD with Sml1 but reduced its interaction with R1-CTD (Fig. 4C). The observed differences were not caused by variation in abundance of the mutant protein (data not shown).

To examine further the functional role of the WE sequence motif, we introduced the WE-to-AD mutation into the full-length RNR1 under the control of the native RNR1 promoter on a centromeric plasmid. The rnr1(WE-to-AD) mutation was lethal to the rnr1Δrnr3Δ double-mutant strain (Fig. 4D). The lethality depended on the presence of the wild-type SML1 gene because removal of SML1 in the rnr1(WE-to-AD)rnr3 double-mutant strain restored cell viability (Fig. 4D). These results suggest that the rnr1(WE-to-AD) mutation may increase the interaction of R1 with Sml1 without significantly compromising R1 function when Sml1 is absent. Interestingly, even in the absence of SML1, the rnr1(WE-to-AD) mutant cell still exhibited a slightly slower growth phenotype relative to cells containing the wild-type RNR1 (Fig. 4D). This growth defect may result from a suboptimal R1 regeneration process caused by a weakened interaction between R1-NTD(WE-to-AD) and R1-CTD, as noted in the yeast two-hybrid experiment (Fig. 4C). In summary, the WE motif of the R1-NTD plays a critical role in mediating the association of R1-NTD with both Sml1 and R1-CTD.

The Sml1 protein is degraded during S phase of the cell cycle and in response to DNA damage (15, 16). We have shown that overexpression of Rnr1 results in an increase in Sml1 level (Fig. 4E), suggesting that increased R1–Sml1 association may stabilize Sml1. To determine whether the increased affinity of the Rnr1(WE-to-AD) mutant protein toward Sml1 might also affect Sml1 stability, we compared Sml1 protein levels between wild-type cells harboring a copy of the rnr1(WE-to-AD) mutant allele with those carrying an extra copy of the wild-type RNR1; both alleles were under the control of the RNR1 promoter. The Sml1 levels from asynchronous cultures showed no obvious difference between the two strains. However, when they were treated with hydroxyurea, the cells containing rnr1(WE-to-AD) exhibited a more stabilized Sml1 level relative to the wild-type control (Fig. 4F). These results suggest that an increase in either the amount of R1 or R1 binding of Sml1 may prevent Sml1 from degradation.

Discussion

The molecular dissection and characterization of domain–domain interactions among Sml1, R1-CTD, and R1-NTD described here, combined with previous knowledge, lead us to propose a competition model for Sml1 inhibition of R1 activity (Fig. 5). In essence, Sml1 and R1-CTD can interact with a shared region in R1-NTD. The R1-CTD of one R1 monomer can be physically accessible to the active site located in the R1-NTD of its neighboring monomer, to achieve regeneration of the active site mediated by the CX₂C motif at the end of the R1-CTD. As a result, occupancy of the shared contact region in R1-NTD by Sml1 precludes R1-CTD from performing R1 regeneration, leading to an inhibition of RNR activity. Alternatively, Sml1 could interfere with substrate binding to or product release from the active site.

Fig. 5.

A model for the role of R1-CTD in regeneration of R1 and its inhibition by Sml1. (a–d) Steps of R1 active-site regeneration by means of an interchain cross-talk between the C terminus of one R1 monomer and the active site of its neighboring monomer. For simplicity, only one active site and one CX₂C motif are highlighted for each pair of monomers. (a) The cysteine residue shown at the top of the R1 active site is converted to a transient thiyl radical through electron relay from the tyrosyl radical located in R2 (8). At the completion of each turnover cycle, a disulfide bond is formed between the conserved redox-active cysteine pair at the active site, thereby inactivating the enzyme (R1_ox, the oxidized form). (b–d) Rereduction of the R1 active site is mediated by the C-terminal CX₂C motif that shuttles reducing equivalents from thioredoxin (Trx) or glutaredoxin (Grx) through two disulfide-exchange steps, resulting in an active enzyme (R1_red, the reduced form). (e) Sml1 competes against R1-CTD in the association with R1-NTD and thus interferes with the regeneration of the active site.

In vitro mutagenesis and kinetic studies support a critical role for the C-terminal cysteine pair (CX₄C or CX₂C) of R1 in regeneration of its active site (21, 36). A detailed mechanistic understanding of how this C-terminal cysteine pair facilitates the disulfide exchange-mediated rereduction reaction has been hindered by the lack of three-dimensional structural information of the C-terminal region of R1 (18, 19). Our results of the S. cerevisiae R1 demonstrate that the C terminus of one monomer suffices to interact directly with the active site of its neighboring monomer in vivo. Although both monomers of R1 are active (37), it is believed that only one of the two monomers of the E. coli R2 homodimer (β₂) or the S. cerevisiae R2 heterodimer (ββ′) is capable of assembling the tyrosyl radical required for catalysis (38–40). Therefore, the asymmetry between the R1 and R2 dimers in the RNR holoenzyme could result in the restricted interaction of the C-terminal CX₂C with only one of the two active sites in the R1 dimer. If this hypothesis proves to be true, then the interchain cross-talk identified here may serve as a predominant mode of active-site regeneration for the RNR enzyme. The interchain cross-talk and asymmetry within the R1 subunit might help explain why R1 forms a dimer or oligomer of dimers.

We have shown that the CI region of the yeast R1 is not essential for the observed cross-talk between the C terminus and the active site of neighboring monomers. Because the C terminus of the CI-deleted R1 mutant protein has a domain structure similar to those of the R1s from E. coli and mammals, we propose that the interchain cross-talk interaction is likely to be a general mechanism for R1 regeneration in other organisms.

Biochemical studies have shown that Sml1 and Rnr1 can associate to form a complex of 1:1 stoichiometry in vitro (14). Our results indicate that the Tyr-688 and Glu-689 residues in R1-NTD are directly involved in the interaction of Rnr1 with R1-CTD and Sml1. The WE-to-AD mutation in R1-NTD compromises its ability to interact with R1-CTD but enhances its interaction with Sml1. The same mutation in full-length RNR1 causes SML1-dependent lethality. The decreased efficiency in rereduction of the active site by R1-CTD in combination with an enhanced inhibition from Sml1 likely underlies the observed cell lethality conferred by the WE-to-AD mutation. These findings illuminate a previously undescribed mode of inhibition of R1 by Sml1.

Our model postulates that Sml1 associates preferentially with the oxidized form of the R1 subunit to interfere with R1 rereduction (Fig. 5). The K_d for binding of Sml1 to the yeast R1 is ∼0.4 μM (14), and that for binding of R1 to R2 is 0.1–0.5 μM (41), all within the physiological concentrations of these proteins (39). The relatively weak and reversible natures of these protein–protein interactions could be of biological significance. For example, the R1 subunit can exist as different oligomeric structures depending on binding of dATP and ATP and their concentrations (7). Similarly, the ability of the R1 and R2 subunits to form the holoenzyme can be regulated by dNTPs and ATP (6). Moreover, the levels of Sml1 and R2 accessible to R1 have been shown to fluctuate during the normal cell cycle and after DNA damage because of changes in protein stability and subcellular localization, respectively (13, 16). All of these factors can contribute to the tight control of RNR activity in vivo.

Although no apparent sequence homolog of Sml1 has been identified in other eukaryotes to date, several lines of evidence suggest that R1 inhibition through interaction with a functional counterpart of Sml1 may be a conserved mode of RNR regulation. First, Sml1 interacts with both mouse and human R1s and inhibits the mouse R1 activity in vitro (14, 17). Second, mutations in SML1 disrupting its binding of the yeast R1 also abolish its interaction with the human R1 (17). Third, the recently identified Schizosaccharomyces pombe Spd1 protein (127 residues) binds R1 and inhibits its activity in vitro (42), suggesting that Spd1 could be a functional equivalent of Sml1 (however, see ref. 11). Further studies into the mechanistic details of R1 modulation by these small protein inhibitors will provide new therapeutic targets for RNR inhibition in treatment of a number of malignancies.

Materials and Methods

Strains and Plasmids.

Growth of yeast strains and genetic manipulations were as described previously (43). 5-Fluoroorotic acid (5-FOA; Sigma, St. Louis, MO) was used at 1 g/liter. The strains used were MHY226 (MATa rnr1::HIS3 ade2 can1 his3 leu2 trp1 ura3 pMH352), MHY784 (MATα rnr1::HIS3 rnr3::kan ade2 can1 his3 leu2 trp1 ura3 pMH352), and MHY802 (MATa rnr1::HIS3 rnr3::kan sml1::kan ade2 can1 his3 leu2 trp1 ura3 pMH352). Yeast two-hybrid analyses were performed in PJ69-4a (MATa trp1-901 leu2-3,112 ura3-52 his3-200 gal4Δ gal80Δ) as described (31). Synchronization of cells in G₁ phase and FACS analyses were as described (44).

A 3.9-kb FspI–KpnI RNR1-containing fragment was subcloned from pSE738 (28) into SmaI–KpnI sites of pRS414 and pRS416 (45), yielding pMH351 and pMH352, respectively, which were used for site-directed mutagenesis to generate desired point mutations and deletions in RNR1. The constructs for yeast two-hybrid assays were derived from vectors pAS2 [Ap^r 2μ TRP1 CYH2 ADH1 promoter-GAL1 (1–147)] and pACT2.2 [Ap^r 2μ LEU2 ADH1 promoter-GAL1 (728–878)], which were described previously (46).

Phenotypic Analyses.

The plasmid shuffle complementation assay (47) was used to determine the effect of RNR1 mutations on yeast cell growth in the strains MHY784 (rnr1Δ rnr3Δ URA3CENRNR1) and MHY802 (rnr1Δ rnr3Δ sml1Δ URA3CENRNR1). Cells bearing both the test RNR1 plasmid (TRP1-containing vector) and endogenous plasmid (URA3 and RNR1-containing vector) were plated on medium containing 5-FOA, which selects for cells that have spontaneously lost the URA3 plasmid harboring the wild-type RNR1. The appearance of 5-FOA-resistant colonies reflected the ability of an RNR1 variant on the TRP1-containing vector to support cell viability in the absence of its wild-type counterpart.

Protein Blotting and Antibodies.

Yeast cells were harvested from early to mid log phase cultures (1–2 × 10⁷ cells per ml). Protein extracts were prepared by using the trichloroacetic acid method and subjected to Western blotting as described previously (48). Monoclonal anti-Myc (9E10) and anti-HA (12CA5) were from Roche Applied Sciences (Indianapolis, IN). Polyclonal anti-Zwf1 (glucose-6-phosphate dehydrogenase) was from Sigma. Polyclonal anti-Gal4 DBD(1–147) was from Covance (Princeton, NJ). Horseradish peroxidase-conjugated goat anti-mouse and goat anti-rabbit antibodies were from Jackson ImmunoResearch, West Grove, PA.

β-Galactosidase Assays.

Cells were grown overnight in appropriate selective liquid medium to stationary phase, diluted 1:100, and grown in fresh medium to early log phase (A₆₀₀ ≈ 0.5). Liquid β-galactosidase assays were performed on chloroform and SDS-permeabilized cells by using the colorimetric substrate o-nitrophenyl-β-galactopyranoside (Sigma) as described (9).

Abbreviations

ACT

activating domain

CI

C-terminal insert

CTD

C-terminal domain

DBD

DNA-binding domain

FL

full-length

5-FOA

5-fluoroorotic acid

NTD

N-terminal domain

RNR

ribonucleotide reductase.