Characterization of 3′hExo, a 3′ Exonuclease Specifically Interacting with the 3′ End of Histone mRNA*

The 3′ end of mammalian histone mRNAs consisting of a conserved stem-loop and a terminal ACCCA interacts with a recently identified human 3′ exonuclease designated 3′hExo. The sequence-specific interaction suggests that 3′hExo may participate in the degradation of histone mRNAs. ERI-1, a Caenorhabditis elegans homologue of 3′hExo, has been implicated in degradation of small interfering RNAs. We introduced a number of mutations to 3′hExo to identify residues required for RNA binding and catalysis. To assure that the introduced mutations specifically target one of these two activities of 3′hExo rather than cause global structural defects, the mutant proteins were tested in parallel for the ability both to bind the stem-loop RNA and to degrade RNA substrates. Our analysis confirms that 3′hExo is a member of the DEDDh family of 3′ exonucleases. Specific binding to the RNA requires the SAP domain and two lysines located immediately to its C terminus. 3′hExo binds with the highest affinity to the wild-type 3′ end of histone mRNA, and any changes to this sequence reduce efficiency of binding. 3′hExo has only residual, if any, 3′ exonuclease activity on DNA substrates and localizes mostly to the cytoplasm, suggesting that in vivo it performs exclusively RNA-specific functions. Efficient degradation of RNA substrates by 3′hExo requires 2′ and 3′ hydroxyl groups at the last nucleotide. 3′hExo removes 3′ overhangs of small interfering RNAs, whereas the double-stranded region is resistant to the enzymatic activity.

The 3 end of mammalian histone mRNAs consisting of a conserved stem-loop and a terminal ACCCA interacts with a recently identified human 3 exonuclease designated 3hExo. The sequence-specific interaction suggests that 3hExo may participate in the degradation of histone mRNAs. ERI-1, a Caenorhabditis elegans homologue of 3hExo, has been implicated in degradation of small interfering RNAs. We introduced a number of mutations to 3hExo to identify residues required for RNA binding and catalysis. To assure that the introduced mutations specifically target one of these two activities of 3hExo rather than cause global structural defects, the mutant proteins were tested in parallel for the ability both to bind the stem-loop RNA and to degrade RNA substrates. Our analysis confirms that 3hExo is a member of the DEDDh family of 3 exonucleases. Specific binding to the RNA requires the SAP domain and two lysines located immediately to its C terminus. 3hExo binds with the highest affinity to the wild-type 3 end of histone mRNA, and any changes to this sequence reduce efficiency of binding. 3hExo has only residual, if any, 3 exonuclease activity on DNA substrates and localizes mostly to the cytoplasm, suggesting that in vivo it performs exclusively RNA-specific functions. Efficient degradation of RNA substrates by 3hExo requires 2 and 3 hydroxyl groups at the last nucleotide. 3hExo removes 3 overhangs of small interfering RNAs, whereas the double-stranded region is resistant to the enzymatic activity.
Mammalian histone mRNAs end with a highly conserved and unique stem-loop structure followed by a single-stranded ACCCA sequence (1). The 3Ј end of histone mRNAs is specifically recognized by two proteins, the stem-loop-binding protein (SLBP) and 3ЈhExo (2). Binding of SLBP requires the nucleotides at the 5Ј side of the stem-loop, whereas binding of 3ЈhExo requires the single-stranded ACCCA at the 3Ј end. In addition, binding of each protein requires specific nucleotides in the stem and the loop. SLBP and 3ЈhExo can bind the 3Ј end of histone mRNA either individually or simultaneously, forming a ternary complex (2). SLBP stimulates binding of 3ЈhExo to the stem-loop and allows 3ЈhExo to form a complex with subopti-mal RNA targets, raising the possibility that the two proteins make direct contact with each other in the ternary complex (2). Binding of SLBP to the stem-loop in histone mRNA precursors is required for formation of the correct 3Ј end of histone mRNAs in the nucleus (3). SLBP bound to the stem-loop accompanies mature histone mRNA to the cytoplasm, where it stimulates histone mRNA translation (4).
The half-life of histone mRNAs is greatly reduced in response to completion or inhibition of DNA replication, resulting in rapid disappearance of histone mRNAs from the cytoplasm and cessation of histone production (5). The stemloop structure is both necessary and sufficient for the selective degradation of histone mRNA and confers the same type of regulation on other mRNAs when introduced at their 3Ј end (6). Sequence-specific and tight binding of 3ЈhExo to the 3Ј end of histone mRNAs simultaneously with SLBP makes this protein a primary candidate for an exonuclease that participates in the rapid degradation of histone mRNAs. However, it has been recently suggested that 3ЈhExo (7) and its putative Caenorhabditis elegans homologue, ERI-1 (8), might play a role in downregulation of RNA interference by degrading small interfering RNAs (siRNAs). 2 Crystallographic studies of the 3Ј exonuclease domain of 3ЈhExo in complex with rAMP demonstrated that 3ЈhExo is a member of a DEDD superfamily of 3Ј exonucleases (9) that includes both RNases and DNases (10). Among the most prominent members of the family are RNase T, poly(A)-specific ribonuclease, the exosome component PM-Scl 100-kDa autoantigen, and the proofreading subunits of DNA polymerases (10). Interestingly, based on data base searches 3ЈhExo and ERI-1 appear to be part of a group of closely related but uncharacterized putative exonucleases present in a number of eukaryotes. To get some insight into the molecular function of 3ЈhExo and its homologues, we carried out experiments to further define the specificity of this enzyme and the regions required for its activity.

EXPERIMENTAL PROCEDURES
Expression and Mutagenesis of 3ЈhExo-The 35 S-labeled 3ЈhExo was synthesized in rabbit reticulocytes using the transcription and translation-coupled system (TNT), as suggested by the manufacturer (Promega). SLBP and the wild-type and mutant forms of 3ЈhExo were overexpressed using the baculovirus expression system (Invitrogen). Briefly, the cDNAs for each protein were cloned into a pFastBac plasmid and used to generate the recombinant viruses. For preparative purification, a 200-ml culture of Sf9 insect cells was infected with each virus and the His-tagged proteins were purified by chromatography on nickel-agarose. Point mutations and internal deletions within the 3ЈhExo coding region were generated using the QuikChange site-directed mutagenesis kit (Stratagene).
RNA Labeling-RNA substrates were labeled at the 5Ј end using T4 polynucleotide kinase (New England Biolabs) and [␥-32 P]ATP. RNAs generated by T7 transcription prior to labeling were treated with calf intestinal phosphatase to remove the 5Ј phosphate. The pre-H2a RNA (50 ng) was labeled at the 3Ј end with [5Ј-32 P]pCp (Amersham Biosciences) using T4 RNA ligase (Ambion) as suggested by the manufacturer.
Partial KOH Hydrolysis of the Pre-H2a RNA-1 ng of the 5Ј labeled pre-H2a RNA (50,000 counts/min) was dissolved in 200 l of 40 mM KOH and incubated at room temperature for 5 min. The solution was neutralized with 5 l of 3 M potassium acetate, pH 5.5, and 20 l of 1 M Tris, pH 8.0. The RNA was precipitated with ethanol and dissolved in 10 l of water.
3ЈhExo Binding Assays-Mobility shift assay was carried out using 25 pmol of baculovirus-expressed 3ЈhExo and/or SLBP and 50 fmol of 5Ј labeled SL RNA in a total volume of 10 l. In the pulldown binding assay, 35 S-labeled 3ЈhExo was bound to either the SLbi RNA or to various H2a RNA mutants pre-annealed to a biotinylated oligonucleotide complementary to the first 17 nucleotides of the RNA. Both assays were carried out as previously described (2).
3ЈhExo Enzymatic Assay-The exonuclease activity of 3ЈhExo was tested using 1, 5, or 25 pmol of baculovirus-expressed protein as indicated for each experiment and 50 fmol of 32 P-labeled RNA substrates in a total volume of 10 l containing of 135 mM KCl, 50 mM Tris, pH 8, 2.5 mM MgCl 2 , 2.5% glycerol, and 1 g/l bovine serum albumin. In some reactions, 10 mM EDTA was used to inhibit the enzymatic activity of 3ЈhExo. The reactions were incubated at 37°C for 30 min, and the RNA was separated in 8% denaturing polyacrylamide gels and detected by autoradiography.

RESULTS
3ЈhExo Prefers the ACCCA at the 3Ј End-We have previously shown that in the presence of EDTA 3ЈhExo can bind to the 3Ј end of histone mRNAs both alone and simultaneously with SLBP (2). Fig. 1A presents a simplified model of how SLBP and 3ЈhExo bind to the target sequence consisting of a highly conserved SL followed by ACCCA. SLBP recognizes primarily the 5Ј side of the SL with an important role played by two adenosines located 2 and 3 nucleotides upstream of the stem, whereas 3ЈhExo binds to the 3Ј side of the structure with an important role played by the ACCCA terminus. We analyzed the role of the ACCCA terminus in the binding of 3ЈhExo in more detail using a pulldown assay as previously described (2). In this assay, 35 S-labeled 3ЈhExo was incubated in the presence of EDTA with 25 pmol of 48-nucleotide wild-type H2a RNA (Fig. 1B) or its various derivatives pre-annealed to a 2ЈO-methyl adaptor oligonucleotide containing biotin at the 3Ј end. The RNA-protein FIGURE 1. 3ЈhExo prefers the ACCCA at the 3Ј end. A, the sequence of the 31-nucleotide stem-loop (SL) structure at the 3Ј end of the mouse H2a-614 histone mRNA and a model depicting the binding of SLBP and 3ЈhExo to this region. B, schematic representation of three major RNA types used to study catalytic and binding activities of 3ЈhExo. The 31-nucleotide SLbi RNA is identical with the SL RNA (panel A) but contains biotin (Bi ) at the 5Ј end. The 48-nucleotide H2a RNA contains 17 additional nucleotides at the 5Ј end that are complementary to an adaptor oligonucleotide containing biotin at the 3Ј end. The 86-nucleotide pre-H2a RNA is identical to the H2a RNA but contains 38 additional nucleotides following the ACCCA. C, the ability of the 48-nucleotide wild-type H2a RNA (lanes 2, 6, and 9) and mutant RNAs (lanes 3-5, 7-8, and 10) pre-annealed to the adaptor oligonucleotide to bind 35 S-labeled 3ЈhExo. The amount of the radioactive 3ЈhExo collected on streptavidin beads in the absence of RNA is shown in lane 1.
complexes were absorbed on streptavidin beads, and the amount of the radioactive protein associated with the RNA determined using SDS/polyacrylamide gel electrophoresis and phosphorimaging. The 48-nucleotide wild-type RNA ending with the ACCCA bound 20 -25% of the input 3ЈhExo (Fig. 1C, lane 2), whereas in the absence of RNA no protein was absorbed on the beads (lane 1). Progressive shortening of the ACCCA region resulted in a strong reduction of binding of 3ЈhExo to the RNA (lanes [3][4][5]. A significant reduction in the efficiency of binding was also caused by extending the 3Ј end with CUAG (ϩ4) or 38 nucleotides (ϩ38), as in the pre-H2a RNA (lanes 7-8). In addition to changing the length of the 3Ј single-stranded tail, we replaced the entire ACCCA sequence with AUUUU. Binding of 3ЈhExo to this RNA was reduced compared with the wild-type RNA by a factor of 20 (lane 10). Therefore, in agreement with our previous results (2), the ACCCA sequence at the end of histone mRNA is optimal for 3ЈhExo binding.
Role of the SAP Domain and the Interdomain Spacer in RNA Binding-The interaction between 3ЈhExo and the 3Ј end of histone mRNA was surprising because 3ЈhExo lacks any known RNA binding domain ( Fig. 2A). The SAP domain (Fig. 2D) has been defined as a 35-residue motif containing an invariant glycine and a conserved distribution of hydrophobic, polar, and bulky amino acids (11). The SAP domain is present in a number of eukaryotic proteins in conjunction with other domains that link these proteins with RNA or DNA metabolism (11). The functional role of the SAP domain is unknown, although it has been suggested that it facilitates binding of proteins to double-stranded DNA (11). In 3ЈhExo, the SAP domain exists together with the 3Ј exonuclease domain ( Fig. 2A). This organization is typical of orthologues of 3ЈhExo in other vertebrates and is also found in C. elegans ERI-1 (T02441), the only known invertebrate protein that contains both the SAP domain and the 3Ј exonuclease domain (8).
We synthesized a 35 S-labeled ⌬SAP mutant lacking the SAP domain (amino acids 76 -110) and analyzed its ability to bind 25 or 100 pmol of the SLbi RNA containing biotin at the 5Ј end (Fig. 1B). No detectable amount of mutant protein was collected on streptavidin beads, demonstrating that deletion of the SAP domain abolished binding of 3ЈhExo to the SLbi RNA (Fig. 2B, lanes 6 and 7). To determine whether deletion of the SAP domain resulted in global misfolding of 3ЈhExo rather than specifically affecting RNA binding, we expressed the ⌬SAP protein in the baculovirus system and tested the ability of the mutant protein to degrade the 5Ј labeled pre-H2a RNA. This 86-nucleotide RNA contains the internally located stem-loop, which prevents complete degradation of the RNA by 3ЈhExo, resulting instead in the formation of a 43-nucleotide product terminating with the stem. This product is 5 nucleotides shorter than the H2a RNA generated in a mouse nuclear extract from the pre-H2a RNA, which contains the ACCCA following the stem-loop (Fig. 2C, lane 5). The ⌬SAP mutant protein at both 5 pmol (Fig.  2C, lane 3) and 1 pmol (not shown) efficiently degraded the RNA substrate to the stem and exhibited enzymatic activity comparable with that of the wild-type 3ЈhExo. This result demonstrates that removal of the SAP domain does not cause gross structural changes in 3ЈhExo and is also consistent with the fact that most nucleases of the DEDD family do not contain the SAP domain and yet display enzymatic activity.
To identify amino acids of the SAP domain critical for RNA binding, we substituted selected amino acids with alanines. The SAP domain of 3ЈhExo contains two positively charged amino acids, Lys 92 and Lys 104 , and two tyrosines, Tyr 109 and Tyr 110 , that are conserved in ERI-1 but not present in other SAP domain proteins (Fig. 2D), suggesting that these residues may participate in sequence-specific RNA binding. Mutant proteins containing alanine at position 92 (K92) or 104 (K104) retained the wild-type efficiency to bind the SLbi RNA in the pulldown assay (Fig. 2D, lanes 5 and 9). No effect on binding was observed when the two mutations were combined in one K92ϩK104 protein (Fig. 2D, lane 11). Alanine substitution of Lys 99 , which is present in nearly all known SAP domains, reduced the ability of 3ЈhExo to bind the SLbi RNA only slightly (Fig. 2D, lane 7). Simultaneous substitution of two tyrosines at positions 109 and 110 with alanines (2Y mutant) had a similar small effect on binding (Fig. 2D, lane 13). Overall, these initial experiments failed to identify residues of the SAP domain that are involved in RNA binding, suggesting that the critical specificity determinants are provided by other amino acids of the domain and/or that simultaneous change of many amino acids may be required to adversely affect RNA binding ability of 3ЈhExo.
Recently, the crystal structure of 3ЈhExo in complex with the SL RNA was solved, and it demonstrated that the SAP domain is indispensable for binding to the RNA and that a crucial role is played by arginine at position 105. 3 Guided by these studies, we substituted this residue with alanine. This single change was sufficient to abolish binding of 3ЈhExo to the SL RNA (Fig. 2E,  lanes 6 and 7), whereas substitution of another arginine in position 96, which according to the crystal structure is not involved in RNA binding, had no effect (not shown).
We next replaced the Lys 111 and Lys 112 with alanines, thus generating the 2K mutant protein. These two lysines are located in the interdomain spacer that separates the SAP domain from the 3Ј exonuclease domain. The 35 S-labeled 2K protein when tested in the pulldown assay did not detectably interact with 25 pmol of the SLbi RNA (Fig. 2D, lane 15). The baculovirus-expressed 2K protein in a wide range of concentrations retained normal enzymatic activity, indicating that the 2K mutation did not affect global folding of the exonuclease (Fig. 2F, lane 4, and not shown).
We also mutated the YYDYI and ELRINEK sequences flanking the exonuclease domain ( Fig. 2A). Mutations within these sequences, including alanine substitutions and partial or complete deletions, significantly reduced or abolished binding to the SLbi RNA (not shown). However, these mutations had also a negative effect on the enzymatic activity of 3ЈhExo (not shown), suggesting that they may partially disrupt the overall structure of the protein rather than specifically affect RNA binding.
Residues of 3ЈhExo Required for the Enzymatic Activity-The 3Ј exonuclease domain of 3ЈhExo is located between amino acids 133 and 311. Based on the amino acid sequence of this domain, we tentatively classified 3ЈhExo as a member of the DEDD family of 3Ј exonucleases (2). All members of this family are characterized by the presence of the four invariant acidic residues required for catalysis (10). The DEDD family has been divided in two subfamilies depending on the presence of either a conserved histidine (DEDDh subfamily) or tyrosine (DEDDy subfamily). Structural studies of the exonuclease domain of 3ЈhExo complexed with rAMP in the presence of Mg 2ϩ classified 3ЈhExo as a member of the DEDDh subfamily and identified the following invariant residues of the signature sequence: Asp 134 , Glu 136 , Asp 234 , Asp 298 , and His 293 (9) (Fig. 2A). The four acidic residues coordinate two magnesium cations and together with histidine at position 293 provide a platform for hydrolytic cleavage of RNA substrates in the 3Ј-5Ј direction (9).
To confirm that the aspartic acids at positions 234 and 298 are required for the exonuclease activity of 3ЈhExo, we substituted each of the two residues with an alanine. This approach was previously used to identify catalytic residues in two other 3Ј exonucleases of the DEDD family (12,13). The two mutant proteins, D234 and D298, were expressed in the baculovirus system and tested for the ability to degrade the 5Ј labeled 86-nucleotide pre-H2a RNA. The wild-type 3ЈhExo (25 pmol) converted all the input RNA into the 43-nucleotide intermediate (Fig. 3A, lane 3), whereas the two mutant proteins at this high protein concentration were inactive, thus confirming the importance of aspartic acids in positions 234 and 298 for catalysis (Fig. 3A, lanes 5 and 7). Human 3ЈhExo and its vertebrate orthologues as well as C. elegans ERI-1 contain another conserved aspartic acid, located in 3ЈhExo at position 198. Mutation of this residue to alanine did not abolish the enzymatic activity of 3ЈhExo, and at 25 pmol the D198 mutant protein displayed a comparable activity to that of the wild-type protein (Fig. 3A, lanes  3 and 4). At lower protein concentrations, the D198 mutant was less active than the wild-type 3ЈhExo (Fig. 4E, lanes 2 and 3). Substitution of a methionine 235 with alanine eliminated the exonuclease activity of 3ЈhExo (Fig. 3A, lane 6). Based on the presence of this methionine, the vertebrate orthologues of 3ЈhExo and C. elegans ERI-1 have been tentatively classified as members of a new DEMD subfamily rather than the DEDD subfamily (8).
We next asked whether the baculovirus-expressed mutant proteins enzymatically inactive because of mutations within the conserved residues retain normal ability to bind the 31-nucleotide SL RNA in the mobility shift assay. The SL RNA was identical to the SLbi but was lacking biotin, thus allowing labeling at the 5Ј end (Fig. 1, A and B). In the presence of 10 mM EDTA both 3ЈhExo and SLBP form a binary complex with the SL RNA, although the interaction of the RNA with 3ЈhExo is not as strong as with SLBP (Fig. 3B, lanes 2 and 3). A ternary complex with the SL RNA is formed in the presence of SLBP and 3ЈhExo (Fig. 3B, lane 4). The same ability to form a binary and a ternary complex in the presence of EDTA was observed for the D234 protein (Fig. 3B, lanes 5 and 6) and D298 protein (not shown, see also Fig. 3C), demonstrating that these two 3ЈhExo mutants retain normal ability to interact with the SL RNA and SLBP. Surprisingly, the M235 protein was unable to interact with the SL RNA or to efficiently form a ternary complex containing SLBP (Fig. 3B, lanes 7 and 8), suggesting that mutation of Met 235 leads to global changes in the 3ЈhExo structure that abolish both binding and catalytic functions of the protein. A significant amount of the SL probe incubated with the M235 was trapped in the well during electrophoresis (not shown), suggesting protein aggregation and further supporting the notion that mutation of methionine 235 results in protein misfolding. We also tested the ability of the ⌬SAP mutant expressed in the baculovirus system to form the binary and ternary complexes. In agreement with the results of the pulldown assay (Fig. 2, B and E), this mutant was unable to interact with the SL RNA and to efficiently form the ternary complex (Fig. 3B, lanes 9 and 10).
SLBP also forms a stable complex with the SL RNA in the presence of 2.5 mM Mg 2ϩ (Fig. 3C, lane 2). As previously shown (2), under these conditions 3ЈhExo degrades the SL RNA, precluding detection of a binary complex (Fig. 3C, lane 3). SLBP prevents degradation of the SL RNA by 3ЈhExo in the presence of Mg 2ϩ , thus allowing detection of a stable ternary complex containing the two proteins (Fig. 3C, lane 4) (2). As expected, the enzymatically inactive D298 protein formed a stable binary complex with the SL RNA also in the presence of magnesium ions and efficiently cooperated with SLBP in forming a ternary complex (Fig. 3C, lanes 5 and 6).
A 3ЈOH Is Required for Efficient Degradation of RNA Substrates by 3ЈhExo-To determine whether 3ЈhExo is sensitive to the nature of the 3Ј end of the substrate, we tested a number of RNAs containing various groups at the 3Ј terminal nucleotide. We generated a ladder of degradation products by treating the 5Ј labeled pre-H2a RNA with either KOH, which leaves a cyclic 2Ј-3Ј phosphate, or nuclease S1, which leaves a 3Ј hydroxyl (14). The products of alkaline hydrolysis were resistant to 3ЈhExo, whereas the full-length RNA present in the same reaction mixture was shortened to the 43-nucleotide intermediate by the enzyme, indicating that the presence of the cyclic phosphate inhibits the 3ЈhExo activity (Fig. 4A, lanes 5 and 6). As expected, the RNA partially degraded by S1 nuclease was efficiently shortened to the 43-nucleotide intermediate (Fig. 4A,  lanes 3 and 4). To further demonstrate that a phosphate group at the 3Ј end prevents enzymatic degradation, we ligated [5Ј-32 P]pCp to the 3Ј end of the pre-H2a RNA, thus generating an 87-nucleotide RNA substrate labeled near the 3Ј end and terminating with a 3Ј phosphate. This RNA was resistant to 25 pmol of 3ЈhExo alone and became degraded by this amount of the enzyme only after simultaneous addition of calf intestinal phosphatase that removes the terminal phosphate (Fig. 4B).
We next tested whether 3ЈhExo can degrade a 5Ј labeled RNA terminating with a 2Ј deoxynucleotide. We used an unstructured 28-mer consisting of 27 ribonucleotides extended at the 3Ј end by a 2Ј deoxycytidine (RNA/2ЈH). The wild-type enzyme at 25 pmol efficiently degraded this substrate, whereas the D298 mutant protein was inactive (Fig. 4C, lanes 2 and 4). In the presence of 1 pmol of the wild-type 3ЈhExo, the degradation of the RNA/2ЈH was very inefficient (Fig. 4C, lane 6), and only a portion of the input was converted to a 23-nucleotide intermediate as determined by using high resolution gel electrophoresis (not shown). The same amount of 3ЈhExo degraded the majority of the pre-H2a RNA to the 43-nucleotide product (Fig. 4E, lane 2), indicating that the presence of 2Ј hydro-gen at the last nucleotide slows down RNA degradation. The reason for the accumulation of the 23-nucleotide RNA is unclear. The presence of two uridines at the 3Ј end of this RNA may indicate that 3ЈhExo at lower concentrations is inefficient in degrading short RNAs when it encounters a tract of uridines. We have previously observed generation of an intermediate ending with uridines during degradation of the stem-loop RNA (2).
In addition to the wild-type 3ЈhExo and the D298 mutant protein we tested the D198 protein. At 1 pmol, this mutant protein compared with the wild-type 3ЈhExo displayed moderately reduced activity on both the RNA/2ЈH and the pre-H2a substrates (Fig. 4C,  lane 7; Fig. 4E, lane 3). We next tested an RNA oligonucleotide identical to the RNA/2ЈH but terminating with a dideoxycytidine (RNA/2Ј3ЈH). This substrate compared with the RNA/2ЈH was more resistant to the wild-type 3ЈhExo, and in the presence of the highest concentration of the enzyme (25 pmol) and 2.5 mM Mg 2ϩ only ϳ30% of the input was completely degraded (Fig. 4D, lane 3). No degradation was observed in the presence of 10 mM EDTA inhibiting the enzyme (Fig. 4E, lane 2). Collectively, these results demonstrate that the nature of the chemical modification at the 2Ј and 3Ј ribose of the last nucleotide greatly affects the ability of 3ЈhExo to degrade RNA substrates and a phosphate exerts a particularly inhibitory effect.
Our earlier studies failed to demonstrate that 3ЈhExo is active on substrates consisting entirely of deoxynucleotides and thus suggested that 3ЈhExo is an RNAspecific exonuclease (2). We have tested a number of additional single-stranded DNA substrates and detected a very limited 3Ј exonuclease activity with some longer oligonucleotides only in the presence of high enzyme concentrations (25 pmol and more). Further studies are required to determine whether 3ЈhExo under certain circumstances can indeed function as DNase.
3ЈhExo Can Remove the 2-Nucleotide Overhangs from siRNAs-Recent genome-wide screening for proteins potentially involved in RNA interference suggested that ERI-1, a C. elegans homologue of 3ЈhExo, may be involved in degrading siRNAs (8). Surprisingly, both 3ЈhExo and ERI-1, when expressed in rabbit reticulocytes, were reported to efficiently degrade double-stranded RNAs and to be inactive on 21-nucleotide single-stranded RNA substrates (8). To test the ability of the baculovirus-expressed 3ЈhExo to degrade siRNAs, we carried out a number of experiments using two 21-nucleotide single-stranded RNAs, siRNA-S and siRNA-A, which upon annealing formed a 19-nucleotide double-stranded region extended on each side by 2-nucleotide 3Ј overhangs. As expected, each 21-mer labeled at the 5Ј end was completely degraded by 25 pmol of 3ЈhExo in the presence of magnesium ions (Fig. 5A, lanes 3 and 6), and no degradation was observed in the presence of EDTA inhibiting the enzyme (lanes 2 and 5). To determine whether formation of the double-stranded RNA prevents degradation by 3ЈhExo, 25 pmol of the enzyme were incubated with 50 fmol of the 5Ј endlabeled siRNA-S either alone, with increasing amounts of the same unlabeled siRNA-S (cold-S), or the complementary siRNA-A (cold-A). The radioactive RNA used in the reaction alone or in the presence of 1 to 10-fold molar excess of the same unlabeled RNA was completely degraded by 3ЈhExo (Fig. 5B, lanes 2-5, degradation products not shown). However, the presence of the siRNA-A efficiently protected the labeled RNA against the activity of 3ЈhExo, leading to accumulation of products 18 -20 nucleotides in length (Fig. 5B,   lanes 6 -8). Increasing the amount of the complementary RNA resulted in more labeled substrate being protected, likely because of increased efficiency in formation of the double-stranded RNA. Generation of an 18-nucleotide product indicates that 3ЈhExo, in addition to removing the 2-nucleotide overhangs, can also remove the first nucleotide of the double-stranded regions. In the presence of 1 pmol of 3ЈhExo the pre-annealed RNA was stable, indicating that the enzyme at this low concentration is unable to remove 3Ј overhangs (Fig. 5C, lanes 5 and 6), although it can partially degrade the single-stranded RNA under these conditions (Fig. 5C, lanes 2-4).
The Intracellular Localization of 3ЈhExo-To determine the intracellular localization of 3ЈhExo, we cloned 3ЈhExo in-frame with C-terminal green fluorescent protein and analyzed the distribution of the fusion protein in HeLa cells using fluorescent microscopy (Fig. 6). The 3ЈhExo-green fluorescent protein fusion protein predominantly accumulated in the cytoplasm and the nucleoli. Localization of the 3ЈhExo-green fluorescent protein to these two compartments was observed in all transfected cells and was independent of the overall level of expression of the fusion protein. This localization is consistent with our fractionation studies, which demonstrated the presence of the endogenous 3ЈhExo both in the cytoplasmic and nuclear extracts (not shown).

DISCUSSION
The highly conserved 26-nucleotide sequence at the 3Ј end of mammalian replication-dependent histone mRNAs containing a 6-base pair stem and a 4-nucleotide loop followed by ACCCA is specifically recognized by two proteins, the stem-loop binding protein (SLBP) and a protein with 3Ј exonuclease activity designated 3ЈhExo (2). Whereas SLBP is known to play multiple roles in biogenesis and metabolism of histone mRNAs (1), the role of 3ЈhExo remains unknown. Here we carried out a number of experiments to further characterize this exonuclease.
Crystallographic studies of the 3Ј exonuclease domain of 3ЈhExo indicate that 3ЈhExo belongs to the DEDDh subfamily of exonucleases and that the invariant residues involved in catalysis include Asp 134 , Glu 136 , Asp 234 , Asp 298 , and His 293 (9). This classification has been confirmed by our mutational analysis demonstrating that alanine substitution of either Asp 234 or Asp 298 eliminates enzymatic activity of 3ЈhExo. The residues of  the DEDDh core identified in 3ЈhExo by the crystallographic studies are highly conserved in all vertebrate orthologues of 3ЈhExo and in other putative exonucleases that share significant similarity with 3ЈhExo, including ERI-1 and three other homologues of this protein in C. elegans (not shown). Interestingly, all these proteins contain a highly conserved aspartate (position 198 in human 3ЈhExo) that is not part of the DEDDh core. Mutation of this residue to alanine did not abolish the enzymatic activity of 3ЈhExo but moderately reduced its ability to degrade RNA substrates.
A structure-based comparison with other known 3Ј exonucleases of the DEDD family revealed that 3ЈhExo is more similar to DNA-specific exonucleases, including a proofreading domain of Escherichia coli DNA polymerase III, than to RNases (9). At least one enzyme of the DEDD family, E. coli RNase T, is both DNase and RNase (10,(15)(16)(17). In fact, RNase T is more active on DNA than on RNA substrates (15) and can complement defects in DNA repair in cells lacking other DNases normally involved in excising damaged nucleotides from DNA (18). This result suggests that participation in some aspects of DNA metabolism may be a natural function of RNase T in addition to its role in RNA metabolism. We demonstrated that 3ЈhExo is primarily an exoribonuclease. Single-stranded DNAs are very poor substrates, if at all, for 3ЈhExo and therefore unlikely to be degraded by this enzyme under physiological conditions. Efficient degradation of RNA substrates by 3ЈhExo requires 2Ј and 3Ј hydroxyl groups at the last nucleotide. The activity of two other DEDD exonucleases, poly(A)-specific ribonuclease (19) and RNase T (16), has been previously shown to depend on the presence of the 3Ј hydroxyl at the last nucleotide of their substrates.
The most characteristic feature of 3ЈhExo is its ability to specifically and tightly bind the stem-loop at the 3Ј end of histone mRNAs. Binding of 3ЈhExo to the stem-loop is facilitated by simultaneous binding of SLBP, which recognizes the opposite platform of the same RNA target. A detailed picture of how 3ЈhExo interacts with the SL RNA and how SLBP facilitates binding of 3ЈhExo to its RNA target will become available through structural studies of the binary and ternary complexes. Deletion of amino acids immediately flanking the 3Ј exonuclease domain had a strong negative effect on RNA binding but at the same time reduced catalytic activity of 3ЈhExo, suggesting that these mutations at least partially affect global folding of the protein rather than target amino acids specifically involved in RNA binding. The strong inhibitory effect on RNA binding ability of 3ЈhExo was caused by deletion of the SAP domain or an alanine substitution of Lys 111 and Lys 112 located immediately to its C terminus. These two mutations did not affect enzymatic activity of 3ЈhExo, suggesting that both the SAP domain and the two lysines are specifically required for interaction with the SL RNA. Indeed, recent x-ray crystallographic studies of the 3ЈhExo-SL RNA complex demonstrated that the SAP domain is involved in binding to the RNA with the crucial role being played by arginine at position 105. 3 Substitution of this residue with alanine was sufficient to abolish binding of 3ЈhExo to the SL RNA.
The sequence-specific interaction of 3ЈhExo with the 3Ј end of histone mRNAs either alone or together with SLBP strongly suggests that 3ЈhExo may play a role in the rapid decay of histone mRNA at the end of S-phase. This function is consistent with the high concentration of 3ЈhExo in the cytoplasm, the site of histone mRNA degradation. ERI-1 has been identified by genome-wide scanning for mutants of C. elegans with enhanced RNA interference, leading to the hypothesis that one function of this exonuclease may be to degrade siRNAs (8). However, as demonstrated recently, ERI-1 exists in a complex with Dicer and, in addition to downregulating the response to exogenous double-stranded RNAs, is required for accumulation of several endogenous siRNAs (20). Based on these studies a new model for the molecular function of ERI-1 has been proposed. In this model, inspired by the unique features of 3ЈhExo, ERI-1 binds to short stem-loops in a group of endogenous RNAs and removes unpaired 3Ј nucleotides, thus generating a structure suitable for synthesis of double-stranded RNA species subsequently cleaved by Dicer and involved in RNA interference (20). Future studies should provide more information on molecular functions of ERI-1 and 3ЈhExo and help to determine whether these two proteins are related functionally in addition to sharing significant sequence similarity.