Mechanism of Ribonuclease III Catalytic Regulation by Serine Phosphorylation

Ribonuclease III (RNase III) is a conserved, gene-regulatory bacterial endonuclease that cleaves double-helical structures in diverse coding and noncoding RNAs. RNase III is subject to multiple levels of control, reflective of its global regulatory functions. Escherichia coli (Ec) RNase III catalytic activity is known to increase during bacteriophage T7 infection, reflecting the expression of the phage-encoded protein kinase, T7PK. However, the mechanism of catalytic enhancement is unknown. This study shows that Ec-RNase III is phosphorylated on serine in vitro by purified T7PK, and identifies the targets as Ser33 and Ser34 in the N-terminal catalytic domain. Kinetic experiments reveal a 5-fold increase in kcat and a 1.4-fold decrease in Km following phosphorylation, providing a 7.4–fold increase in catalytic efficiency. Phosphorylation does not change the rate of substrate cleavage under single-turnover conditions, indicating that phosphorylation enhances product release, which also is the rate-limiting step in the steady-state. Molecular dynamics simulations provide a mechanism for facilitated product release, in which the Ser33 phosphomonoester forms a salt bridge with the Arg95 guanidinium group, thereby weakening RNase III engagement of product. The simulations also show why glutamic acid substitution at either serine does not confer enhancement, thus underscoring the specific requirement for a phosphomonoester.

A number of bacteriophages employ RNase III in their infection strategies. The coliphage T7 expresses multiple transcripts whose maturation involves RNase III action. Immediately following T7 infection of Escherichia coli, the host RNA polymerase transcribes the phage early region, generating a ~7,000 nt polycistronic mRNA precursor that is processed co-transcriptionally by RNase III at five sites to provide the mature, optimally functional mono-and dicistronic mRNAs 15,16 . One of the early mRNAs encodes the single-subunit RNA polymerase that transcribes the T7 middle and late genes, yielding large amounts of mRNA precursors that also are processed by RNase III 16 . While RNase III is not essential for T7 growth under standard growth conditions, it is important for optimal translation of at least several phage mRNAs 17 . RNase III is a low abundance enzyme 11 , and no new RNase III is synthesized during T7 infection since host transcription is suppressed. Thus, the copious production of T7 mRNAs is expected to place a premium on the ability of the existing RNase III population to function at maximum efficiency.
The T7 0.7 early gene encodes a ~40 kDa polypeptide that has an N-terminal domain (~30 kDa) possessing an ATP-dependent, serine/threonine-specific protein phosphotransferase activity (T7PK), and a C-terminal domain (~10 kDa) that exerts host transcription shut-off [18][19][20] . Expression of T7PK results in the phosphorylation of multiple cellular proteins, including translation factors and the β ′ subunit of the host RNA polymerase [21][22][23] . Phosphorylation of the β ′ subunit renders RNA polymerase responsive to otherwise weak transcription terminator signals, thereby attenuating host gene expression 23 . T7PK undergoes self-phosphorylation on serine, which downregulates its catalytic activity 19,20,24 . T7PK is not essential for T7 growth under normal laboratory conditions, but is practically essential at elevated temperatures or during carbon/energy limitation 25 . The T7PK-dependent modification of components of the gene expression machinery suggests a primary role of T7PK in redirecting gene expression from host to phage as part of the infection strategy.
Several E. coli enzymes involved in T7 RNA metabolism are modified during infection. The phosphorylation of RNA degradosome components RNase E and RNA helicase RhlB serves to stabilize the mRNAs synthesized by the T7 RNA polymerase 26,27 . RNase III also is phosphorylated during T7 infection in a T7PK-dependent manner, and the catalytic activity of the enzyme increases ~4-fold upon phosphorylation in vivo 28 . Serine was identified as the in vivo target; however, the location of the serine(s) was not determined, and how phosphorylation enhances catalytic activity was not defined. We present here an analysis of T7PK phosphorylation of RNase III, and describe the structural basis for the catalytic enhancement of dsRNA processing by phosphorylation.

Results and Discussion
T7PK phosphorylates serine in the α 2 -α 3 loop of the RNase III catalytic domain. The observations that E. coli (Ec) RNase III is phosphorylated on serine in T7-infected cells 22,28 , and that the N-terminal catalytic domain (RIIID) (encompassing residues 1-150) is phosphorylated by T7PK in vitro 24 afforded a strategy for identifying the phosphorylation site(s), using the RIIID polypeptide as the target for selective mutation. The Ec-RIIID polypeptide contains ten serines (Fig. 1). Based on a homology-modeled Ec-RNase III structure 29 , six of the ten serines (Ser31, 33, 34, 103, 127 and 148) are located on the protein surface (Suppl. Figs S1 and S2), and were The regions highlighted in blue correspond to the RNA-binding motifs 3 and 4 (RBM3 and RBM4) as described elsewhere 37 . Supplementary Figure S1 provides the alignment of complete RNase III polypeptide sequences.
Scientific RepoRts | 6:25448 | DOI: 10.1038/srep25448 chosen as targets for alanine mutation. The proximity of S31, S33 and S34 in the loop connecting the α 2 and α 3 helices ( Fig. 1) also prompted the creation of double mutations within the loop. The mutant RIIID polypeptides were purified as soluble, N-terminal hexahistidine (H6)-tagged species and in vitro phosphorylation assays were performed as described (see Materials and methods) using purified recombinant T7PK 24 . Assays of the S31A, S33A, S34A, S103A, S127A, and S148A mutants are shown in Fig. 2. The reaction involving RIIID (Fig. 2a, lanes 2 and 7) contains a 32 P-labeled species that migrates with a mass commensurate with the size of the H6-tagged RIIID polypeptide (~20 kDa), and confirms that RIIID is a substrate for T7PK, as reported elsewhere 24 . Selfphosphorylated T7PK (~30 kDa) is the second, slower-migrating species 24 . The S31A, S103A, S127A and S148A RIIID mutants are phosphorylated to a comparable extent as RIIID (Fig. 2a, lanes 3, and 8-10). In contrast, the S33A and S34A mutants exhibit strongly reduced phosphorylation (Fig. 2a, compare lanes 4 and 5 with lane 2), while the S33A, S34A double mutant is essentially fully resistant to phosphorylation (Fig. 2b, compare lane 4 with lane 2; see also Table 1).
In order to rule out any indirect effect of the C-terminal dsRBD on phosphorylation we also assessed the impact of the S33A, S34A double mutation on phosphorylation of full-length enzyme. Since the electrophoretic mobilities of T7PK and RNase III are similar, a quantitative analysis was not possible. However, the S33A,S34A double mutation also suppresses phosphorylation of the RNase III polypeptide (Fig. 2c, compare lanes 3 and 5 with lane 2). We conclude that (i) S33 and S34 are targets in the full-length enzyme, and (ii) the dsRBD is not a phosphorylation target. In this regard, alanine mutation of the two surface-located serines in the dsRBD (S195 and S198) (see Suppl. Figs S1 and S2) did not alter phosphorylation (Fig. 2c, lane 4). Interestingly, the RIIID S31A, S33A mutant is phosphorylated to a comparable extent as RIIID (Fig. 2b, compare lanes 2 and 3). The enhancement of S34 phosphorylation by the S31A mutation may reflect a conformational change in the α 2 -α 3 loop that alters the T7PK-RNase III interaction (see also below). T7PK self-phosphorylation is significantly reduced in the reaction involving the S33A, S34A RIIID double mutant (Fig. 2b, compare lanes 4 and 2), and to a lesser extent with S34A (Fig. 2a, compare lanes 5 and 2). The reduction is not due to differing amounts of protein (see lower While the mutational approach shows that the α 2 -α 3 loop containing S33 and S34 (hereafter termed the phospholoop -see Fig. 1) is the target for T7PK action, it cannot distinguish between the possibilities that (i) S33 and S34 are phosphorylated in a stochastic, mutually exclusive manner, or (ii) that only one of the two serines is the target, with the other serine important for target recognition and/or reactivity. Thus, several attempts were made to map the Ec-RNase III phosphorylation site(s) by mass spectrometric analysis of enzymatically-generated peptides. However, the analyses were inconclusive, as the tryptic pentapeptide [(R)SASSK 35 ] containing S33 and S34 (underlined boldface) was the only member of the total set of tryptic peptides that was not represented, in either phosphorylated or unphosphorylated form, in a coverage map of the Ec-RNase III polypeptide (S. G. and A.W.N., unpublished results). The failure may reflect the short length of the peptide; however, attempts to map the site by using mutation to create longer tryptic peptides also were unsuccessful (S.G. and A.W.N., unpublished results).
The stoichiometry of RIIID phosphorylation was determined by direct radiometric measurement of the 32 P-labeled polypeptide in excised polyacrylamide gel slices (see Materials and methods). The analysis reveals 0.32 mole phosphate incorporated per mole RIIID homodimer (Table 1). In contrast, the S33A,S34A RIIID double mutant exhibited a level of phosphorylation no greater than the background value (0.04 mol phosphate per mol protein). Additional treatment of phosphorylated RIIID with fresh T7PK and ATP did not increase the level of incorporation, indicating that a fraction of RIIID is resistant to phosphorylation.

Phosphorylation increases the k cat for RNase III cleavage of substrate.
To determine whether phosphorylation enhances Ec-RNase III catalytic activity in vitro we carried out a kinetic analysis of phosphorylated and unphosphorylated enzyme. The phosphorylation reaction conditions allowed maximal covalent modification (~0.32 mol phosphate per RNase III homodimer -see Table 1). The control (mock phosphorylation) reaction contained all components except for Mg 2+ ion, which is required for T7PK catalytic activity 19,20 . As a control, the same kinetic analysis was performed for the S33A, S34A mutant, which is fully resistant to phosphorylation (see Fig. 2b and Table 1). The treated enzymes were purified (see Materials and methods) and employed in multiple-turnover cleavage assays using internally 32 P-labeled R1.1 RNA as substrate. R1.1 RNA (Fig. 3a) has been used in previous enzyme kinetic analyses [30][31][32] and is cleaved by Ec-RNase III at a single phosphodiester, forming 47 nt and 13 nt products. A time course assay of R1.1 RNA cleavage (Fig. 3b,c) shows that phosphorylated RNase III cleaves substrate more rapidly than mock-phosphorylated enzyme. Moreover, the catalytic action of the S33A, S34A double mutant (Suppl. Fig. S3) is not increased by T7PK treatment, thus providing additional support that the α 2 -α 3 loop is the phosphorylation target. The steady-state parameters k cat and K m ( Table 2) were determined from the data in Figure 3d, and show that phosphorylated RNase III exhibits a 5-fold greater k cat and a 1.4-fold lower K m compared to unphosphorylated enzyme. Together, the values provide a 7.4-fold increase in catalytic efficiency (k cat /K m ). Since only ~30% phosphorylation of RNase III was achieved, the 7.4-fold increase most likely represents a lower limit to catalytic efficiency enhancement.
A gel mobility shift assay was used to assess whether phosphorylation alters substrate affinity. In this assay Mg 2+ was replaced by Ca 2+ , which prevents phosphodiester cleavage while supporting substrate binding 33 . Single complexes with similar mobilities were observed for phosphorylated and mock-phosphorylated enzyme, and complex formation exhibited closely similar concentration dependencies (Suppl. Fig. S4). This result is consistent with the minor change in K m (see above) and provides further support that phosphorylation primarily affects one or more events following substrate binding.
Evidence for enhanced product release and specific requirement for a phosphomonoester. A proposed kinetic scheme for RNase III involves (i) substrate recognition by the dsRBD; (ii) substrate engagement by the RIIID; (iii) phosphodiester hydrolysis; and (iv) product release 6,7 . Based on the gel shift assay results (Suppl. Fig. S4) and only a minor change in K m (Table 2), an effect of phosphorylation on step (i) or (ii) can be ruled out. The increase in k cat , however, suggests a change in the rate of phosphodiester hydrolysis (iii) and/or product release (iv). To differentiate between these two possibilities, time-course assays of R1.1 RNA cleavage were performed under single-turnover (enzyme excess) conditions, such that the measured rate (described by a single exponential decay constant, k 2 ) would only reflect the hydrolytic step (see Suppl. Fig. S5). The calculated k 2 values ( and mock-phosphorylated Ec-RNase III. Based on the increased k cat , but unaltered k 2 (Table 2), we conclude that the step responsive to phosphorylation is product release.
To determine whether the rate enhancement reflects introduction of negative charge in the α 2 -α 3 phospholoop, the catalytic activity of the S33E, S34E double mutant enzyme was compared with wild-type enzyme. Similar to a serine phosphomonoester at pH 7.5, two glutamic acid residues would provide two negative charges 34,35 . If negative charge is the functionally important factor, then the double mutant protein would exhibit enhanced activity compared to the normal enzyme. In a time-course assay under conditions of excess substrate the S33E, S34E mutant cleaved R1.1 RNA with an efficiency comparable to, but not greater than unphosphorylated RNase III (S.G. and A.W.N., data not shown). We conclude that the catalytic enhancement reflects a specific requirement for a phosphomonoester group in the α 2 -α 3 phospholoop, rather than negative charge per se.
Computational approaches suggest a structural model for phosphorylation dependent catalytic enhancement. Computational modeling was used to gain insight on how α 2 -α 3 loop phosphorylation enhances catalytic efficiency, and the specific requirement for a serine phosphomonoester. A homology model of Ec-RNase III 29 was used (Fig. 4) that contained a cleaved dsRNA substrate. This structure probably corresponds to an intermediate in the product release pathway, immediately following phosphodiester hydrolysis 36 . To assess local conformational changes as a consequence of S33 or S34 phosphorylation, molecular dynamics (MD)   Table 2. Effect of phosphorylation on the kinetic parameters of substrate cleavage by Ec-RNase III. Kinetic parameters were determined using internally 32 P-labeled R1.1 RNA as substrate (see Materials and methods). RNase III(+ P) refers to Ec-RNase III subjected to phosphorylation by T7PK and ATP, followed by purification (see Materials and methods). RNase III (−P) refers to a mock phosphorylation reaction where MgCl 2 was omitted from an otherwise complete reaction. Experiments were performed in duplicate. The K m and k cat values were determined by nonlinear least-squares curve-fitting (Kaleidagraph v3.5) of the data points (Fig. 3d) to a Michaelis-Menten scheme. The k cat value was calculated by dividing V max by the enzyme concentration. The standard errors are provided. The k 2 value is the exponential decay constant for R1.1 RNA cleavage under single-turnover conditions. To determine the k 2 values, reactions were performed as described in Materials and methods and the exponential decay constant obtained by fitting the time-course reaction progress curve to the single exponential equation, y = m 1 + m 2 · e −m3t , where t is time (min) and m 3 is the decay constant k 2 (min −1 ). The k 2 values are the average of two experiments, and the maximum error values are provided.
simulations were performed on Ec-RNase III and its two monophosphorylated forms (pS33 or pS34). Figure 5a and Supplemental Figure S8 show that the S33 phosphomonoester is engaged in a bidentate salt bridge with the R95 guanidinium group, and also a more flexible (i.e. weaker) ionic interaction with the K35 side chain. In contrast, the S34 phosphomonoester interacts primarily with K35 (Fig. 5b), with a salt bridge to R95 only infrequently observed (Suppl. Fig. S8, and Suppl. Tables S3 and S4). Simulations of the diphosphorylated (pS33 and pS34) RNase III-product complex exhibit the same pS33-R95 interaction (Suppl. Fig. S9) seen in the monophosphorylated (pS33) complex (Fig. 5a), providing further support for a dominant effect of S33 phosphorylation on repositioning the R95 side chain. R95 is a feature of RNA-binding motif 4 (RBM4) (see Fig. 1) which interacts with the 2 bp distal box (db) that is present in RNase III substrates >11 bp in length 37 . The db (see Fig. 3a) is important for substrate reactivity, as its deletion or mutation causes a loss of binding affinity 38 . In this regard, the dsRNAs used in the simulations are 9 bp in length with a 2 nt, 3′-overhang at each end, and do not contain a complete db (see Suppl. Fig. S2). Thus, the modeling would not be expected to show a direct R95 side chain interaction with the RNA. Moreover, other crystal structures of Aa-RNase III bound to cleaved RNAs exhibit RBM4-db interactions that are dependent upon RNA structure and the specific type of product complex (Suppl . Table S7). Furthermore, these observations may relate to the variability of RBM4 sequence and length among RNase III orthologs 37,38 . Despite the limitations of the homology-modeled structure, the S33 phosphorylation-dependent repositioning of the R95 side chain suggests a mechanism for facilitated product release, in which the ionic interaction of the R95 side chain with the substrate distal box is replaced by a bidentate ionic interaction with the S33 phosphomonoester.
The modeling also explains why glutamic acid substitution at position 33 or 34 does not confer a rate enhancement. The predicted interaction of R95 with a glutamic acid at position 33 (Suppl. Fig. S11a) is significantly weaker than the interaction involving the S33 phosphomonoester (Suppl. Fig. S12 and Suppl. Tables S3 and S4). Similarly, the interaction of K35 with E34 (Suppl. Fig. S11b) is less stable than with pS34. Moreover, MD simulations of the S33E, S34E double mutant (Suppl. Fig. S11c), which formally provides the same double-negative charge as a single  S33 or S34 phosphomonoester, indicate that an additional acidic residue at position 34 does not provide a stabilized interaction with R95 (see Suppl. Fig. S12 and Suppl. Tables S3 and S4). In contrast to the bidentate pS33-R95 side chain interaction, the observed salt bridge consists of a monodentate engagement of R95 with the E33 side chain, and no involvement of the E34 side chain. We also note that Glu-Arg and Glu-Lys interactions generally are ~1-2 kcal/ mol weaker than pSer-Arg or pSer-Lys interactions 39 . Hence, we conclude that the phosphomonoester group offers optimal geometric and electronic features that allow stable bidentate salt bridge formation with the R95 guanidinium group, and that the resulting weakened interaction between RBM4 and the distal box facilitates product release.
In the proposed model, serine phosphorylation also could be expected to weaken the binding of substrate as well as product. If so, this could counteract any rate enhancement provided by facilitated product release. However, we find that phosphorylation has only a minor effect on the stability of the enzyme-substrate complex (see Fig. S4). While a definitive explanation of how phosphorylation selectively acts on product release is not yet available, it should be noted that the product complex formally differs from the substrate complex by the presence of a doubly-negatively-charged phosphomonoester. The phosphomonoester may provide additional electrostatic destabilization of the product, compared to the substrate complex, and in turn could confer a greater sensitivity of the RBM4-db interaction to serine phosphorylation.
Bioinformatics analysis of the α 2 -α 3 phospholoop. Alignment of RNase III sequences from bacteria that are known hosts for T7-related phages (Suppl. Fig. S16) reveals a conservation of the α 2 -α 3 phospholoop and R95. However, the overall sequence identity of the RNase III orthologs (Suppl . Table S9) is >90% if Stenotrophomonas and Vibrio are excluded, so regions in addition to the α 2 -α 3 loop also are highly conserved. In contrast, a broadened alignment of RNase III sequences, including those from bacteria for which there is no evidence for a T7-related phage, shows an α 2 -α 3 loop of variable length and a lower conservation of S33 and S34 (Suppl. Fig. S17). Why would a bacterial cell have phosphorylation sites that benefit phage reproduction? It is possible that T7PK may recognize a pre-existing regulatory network involving a cell-encoded protein kinase. A GenBank database search using T7PK as query failed to identify a putative cellular ortholog. However, it is possible that strong selective pressure on the phage genome may have caused significant sequence divergence from an ancestral, host-derived kinase gene. Alternatively, convergent evolution could have produced a phage protein kinase with similar target specificity.

Materials and Methods
Chemicals and reagents were molecular biology grade and were obtained from Sigma-Aldrich or Thermo Fisher Scientific. Standardized 1M MgCl 2 was from Sigma-Aldrich. Ribonucleoside 5′-triphosphates were obtained from Roche Molecular Biochemicals. The radiolabeled nucleotides [α -32 P]UTP (3000 Ci/mmol) and [γ -32 P] ATP (3000 Ci/mmol) were purchased from GE Healthcare Life Sciences. Lambda phage protein phosphatase and T4 polynucleotide kinase were purchased from New England BioLabs. Dialysis membranes (Spectra-Por CE 3500, 10,000 MWCO) were purchased from Thermo Fisher Scientific. Ni 2+ -NTA affinity chromatography resin, biotinylated thrombin and streptavidin-agarose were purchased from EMD Millipore. Protein assay kits and SDS-PAGE protein standards (low MW range) were from Bio-Rad Laboratories. ICON TM concentrators, agarose, and NuPAGE Precast Bis-Tris or Tricine gels (12% and 15%, respectively) were purchased from Thermo Fisher. Oligodeoxynucleotides used for mutagenesis and for in vitro RNA synthesis were provided by Thermo Fisher in fully deprotected form, and were purified by denaturing gel electrophoresis, then stored at − 80 °C in 10 mM Tris, 1 mM EDTA (TE) buffer (pH 8.0) until further use. RNA synthesis. Internally 32 P-labeled R1.1 RNA was synthesized in vitro using T7 RNA polymerase (purified in-house as described 40 ) and an oligodeoxynucleotide template essentially as described 41 , using [α -32 P]UTP (100 Ci/mol) as radiolabel. The concentration of all four rNTPs was 1 mM, and 400 units of T7 RNA polymerase was used, in 0.1 ml reactions. RNA was purified by electrophoresis in a 15% polyacrylamide gel containing 7M urea in Tris/Borate/EDTA, and stored at − 20 °C in TE buffer (pH 7.0).

Mutagenesis and protein purification.
Site-directed mutagenesis was accomplished using QuikChange multisite-directed mutagenesis kits (Agilent Technologies) and a single mutagenic oligodeoxynucleotide (sequences available on request). Ec-RNase III and mutant versions were overproduced in H6-tagged form using E. coli BL21(DE3)recA,rnc105 cells 41 that carried the respective recombinant pET-15b plasmid, and purified by affinity chromatography essentially as described 41 . Briefly summarized, cells were grown with vigorous aeration to mid-log phase at 37 °C in LB media containing ampicillin. Approximately 4 hours following addition of IPTG (1 mM final concentration), cells were collected by centrifugation at 4 °C. The pellet was resuspended in column loading buffer, sonicated until lysis was complete, then centrifuged and the clarified supernatant applied to a Ni 2+ -NTA column. Following repeated column washes, the fractions containing the eluted protein were collected and dialysed against 1 M NaCl, 60 mM Tris-HCl (pH 7.9), 1 mM EDTA, 1 mM DTT for 12 hr at 4 °C (EDTA and DTT can be omitted). An equal volume of glycerol was added and the protein stored at − 20 °C. The H6-tag was removed as needed by treatment with biotinylated thrombin as described 24 .
T7PK phosphorylation of protein in vitro. T7 protein kinase (T7PK) was purified in H6-tagged form and dephosphorylated as described 24 . A standard phosphorylation assay involved an initial incubation of equimolar amounts of RNase III (or RIIID) and dephosphorylated T7PK for 5 min at 30 °C in 2 mM NH 4 Cl, 30 mM Tris-HCl (pH 7.2), 1 mM DTT, 0.1 mM EDTA. Then, MgCl 2 was added (15 mM final concentration), followed by [γ 32 -P]ATP (12 Ci/mol; 1 mM final concentration). Following a 10 min at 30 °C, an additional aliquot of T7PK was added, at half the original amount. After a 5 min incubation excess EDTA (20 mM final concentration) was added. Aliquots were combined with SDS-PAGE loading buffer, heated at 100 °C for 3 min and analyzed by SDS-PAGE. Proteins were visualized by Coomassie Brilliant Blue R staining and 32 P-radioactivity detected by phosphorimaging (Typhoon 9400 System). To prepare phosphorylated nonradioactive RNase III for use in RNA cleavage and binding assays, phosphorylation reactions were performed using RNase III with the H6-tag enzymatically removed by thrombin treatment 24 , dephosphorylated T7PK, and a 1 mM ATP concentration. The reaction mixture was loaded onto a Ni +2 -NTA spin column (Qiagen) and centrifuged at 500 × g for 5 min. Phosphorylated RNase III free from H6-T7PK was recovered in the eluted volume, dialysed against storage buffer and stored at a concentration of ~0.5 mg/ml in 50% glycerol (v/v) at − 20 °C until further use.
Substrate cleavage assay. Substrate cleavage assays were performed essentially as described 41 using internally 32 P-labeled R1.1 RNA (see above, and Fig. 3) and purified WT or mutant H6-tagged RNase III. Briefly, assays were conducted by incubating R1.1 RNA and RNase III (specific amounts indicated in the relevant Figure and Table legends) in 160 mM NaCl, 30 mM Tris-HCl (pH 8) for 1 min at 37 °C. Reactions were initiated by adding MgCl 2 (10 mM final concentration, unless otherwise specified), followed by incubation at 37 °C for the specified time. Aliquots were removed and combined with gel loading dye containing excess EDTA, then analyzed by electrophoresis in 15% (w/v) polyacrylamide gels containing 7 M urea and TBE buffer 41 . RNA was visualized by phosphorimaging (Typhoon 9400 system) and quantified using ImageQuant software. Curve fitting for kinetic parameter determination used Kaleidagraph software (v.3.5) (see also Table 2 legend). Single-turnover cleavage assay conditions are described in the legend to Supplemental Figure S5.
Computational modeling. A homology model of Ec-RNase III was built as described 29 , and was based on the crystal structure of Aquifex aeolicus (Aa) RNase III (PDB entry 2NUG) 36 . The Aa-RNase III structure corresponds to a complex of the enzyme with the minimal product of dsRNA processing 37 , consisting of a 9 bp dsRNA with a 2-nt, 3′-overhang at each end, and is proposed to represent an early step in the product release pathway 36 . The product-bound homology model of Ec-RNase III was simulated in different states: (i) wild-type homodimer (wt-wt′); (ii) wild-type with both subunits phosphorylated at Ser33 (pS33-pS33′); (iii) wild-type with both subunits phosphorylated at Ser34 (pS34-pS34′); (iv) wild-type with both subunits diphosphorylated at Ser33 and Ser34 (pS33, pS34-pS33′, pS34′); (v) the S33E mutant-mutant homodimer (S33E-S33′E); (vi) the S34E mutant-mutant homodimer (S34E-S34′E); and (vii) the S33E, S34E double mutant homodimer (S33E, S34E-S33′E, S34′E). For each of the simulated systems, two independent 50 ns molecular dynamics (MD) simulations were performed in order to refine the homology models and explore possible local conformational changes due to phosphorylation or mutation. In the monophosphorylated states (pS33 or pS34) phosphoserine was modeled in doubly unprotonated form (i.e. with a − 2 charge), since the first and second pKa values of the phosphomonoester group are 2.19 and 5.78, respectively (see ref. 42 and references therein). In the diphosphorylated state (pS33 and pS34), two possible protonation states were considered: doubly unprotonated (− 2 charge) pS33 together with singly unprotonated (−1 charge) pS34, or vice versa. As a control for the homology modeling protocol, and taking into account that Aa-RNase III is also phosphorylated in vitro by T7PK (Suppl. Fig. S6), MD simulations were also performed for Aa-RNase III in complex with a minimal product dsRNA (PDB entry 2NUG) 36 , which is the same structure used as a template in the homology modeling. Phosphorylation was only considered at Aa-S31, i.e. the residue equivalent to Ec-S33 (see Fig. 1). The non-phosphorylated and phosphorylated forms of the Aa-RNase III enzyme were simulated for 100 ns each. All the MD simulations were performed using the NAMD program 43 , with additional details available in the Supplementary Information (SI). The computational images were generated with the programs ESPript 44 (Fig. 1) and VMD 45 (Figs 4 and 5).