Action of RNase II and Polynucleotide Phosphorylase against RNAs Containing Stem-Loops of Defined Structure

Abstract

The 3′→5′ exoribonucleases, RNase II and polynucleotide phosphorylase (PNPase), play an essential role in degrading fragments of mRNA generated by prior cleavages by endonucleases. We have assessed the ability of small RNA substrates containing defined stem-loop structures and variable 3′ extensions to impede the exonucleolytic activity of these enzymes. We find that stem-loops containing five G-C base pairs do not block either enzyme; in contrast, more stable stem-loops of 7, 9, or 11 bp block the processive action of both enzymes. Under conditions where enzyme activity is limiting, both enzymes stall and dissociate from their substrates six to nine residues, on average, from the base of a stable stem-loop structure. Our data provide a clear mechanistic explanation for the previous observation that RNase II and PNPase behave as functionally redundant.

In the bacterium Escherichia coli, degradation of mRNA is almost always initiated by an endoribonuclease, usually RNase E (1, 26). At least two 3′ exoribonucleases subsequently attack the newly created 3′ termini generated by RNase E. One of these, RNase II, a monomer with a molecular mass of 72.5 kDa, is hydrolytic (31) and accounts for up to 90% of the exoribonucleolytic activity in crude extracts (12). The other, polynucleotide phosphorylase (PNPase), a trimer with 78-kDa subunits, is phosphorolytic and accounts for the remaining 10% of the exoribonuclease activity in E. coli extracts (12). Although strains singly mutated in the genes encoding RNase II (rnb) or PNPase (pnp) exhibit a mild phenotype, double mutants deficient in both PNPase and RNase II are inviable (13). This finding has been interpreted to indicate that these exonucleases are functionally redundant but collectively essential. Other data, however, suggest that RNase II and PNPase are not functionally equivalent but are differentially sensitive to RNA secondary structure (3, 7, 8, 16, 19, 24, 28). In such cases, PNPase is required for the degradation of highly structured RNAs and RNase II cannot substitute (7, 8, 19). Moreover, RNase II, but not PNPase, may actually stabilize some RNAs (3, 28). In addition, while RNase II behaves as a soluble monomeric enzyme (31), PNPase can be assembled into a multienzyme complex, the degradosome (5, 27, 29). In the complex, PNPase can degrade extensively structured RNA substrates in concert with RhlB, a putative DEAD-box RNA helicase (10, 29). Alternatively, the action of PNPase against folded RNAs can be stimulated by prior 3′ polyadenylation of such substrates (2, 33). RNase II can also be stimulated by polyadenylation in vitro, but to a more limited extent (7).

In order to resolve the paradoxical properties of RNase II and PNPase, we compared their abilities to degrade short synthetic RNA substrates containing a single stem-loop of defined size and thermal stability. This would permit us to test directly whether both enzymes are functionally equivalent. In addition, we could measure the minimum size of base-paired stems which would stall each enzyme and consequently predict which natural secondary structures are intrinsically sensitive to RNase II and which would require PNPase and/or RNA helicases for their degradation.

MATERIALS AND METHODS

Synthesis of RNA in vitro.

Both strands of the DNA templates for the stem-loop structures shown in Fig. 1a containing KpnI and BamHI cohesive termini at their 5′ and 3′ ends, respectively, were synthesized at the NAPS Unit, University of British Columbia, annealed, and ligated into the vector pTZ18U (25) by using standard methods (30). Appropriate recombinants, verified by DNA sequence analysis, were modified further by ligation of annealed oligonucleotides GC-10 (5′-GATCC[A]₃₀T) and GC-11 (5′-CTAGA[T]₃₀G) between the BamHI and XbaI sites to generate templates for the SLxA RNAs. For internally labeled RNAs, transcription was performed essentially as previously described (11) by using 1.5 pmol of DNA template linearized with either BamHI (SLx RNAs), XbaI (SLxA RNAs), or HindIII (SLxR RNAs), 25 μCi of [α-³²P]CTP, and 20 U of T7 RNA polymerase (Promega) but with the inclusion of 0.18 U of pyrophosphatase (Sigma) to enhance yields. For 5′-labeled RNAs, the GTP concentration was reduced to 125 μM and 25 μCi of [γ-³²P]GTP was substituted for [α-³²P]CTP. RNA structure mapping was performed by using 5′ [γ-³²P]-labeled substrates as described previously (21, 22).

FIG. 1.

Schematic diagram of the structures of RNA substrates. (a) Three different classes of 3′ extensions, denoted as SLx, where x is the number of base pairs in the stem-loop. The common 5′ end of all substrates is 5′ pppGGGAAUUCGAGCUCGGUAC. An imperfect inverted repeat in the SLxR RNAs (see the text) is shown by arrows. (b) Primary and predicted secondary structures of the four stem-loops examined in this work.

Exonuclease assays.

RNase II (6) and PNPase (8) were purified to at least 90% homogeneity as described previously. Assays of RNase II were performed essentially as previously described (6) in a buffer containing 17 mM HEPES-KOH (pH 7.5), 100 mM KCl, 1 mM MgCl₂, 2 mM dithiothreitol, and 5% glycerol with 2 pmol of RNA and 0.1 ng of purified RNase II (0.4 milli-units; 1.4 fmol) in 25 μl at 37°C for the times noted in the figures. Assays of PNPase (0.2 ng [0.9 fmol]; 8 × 10⁻⁶ U) were performed similarly in a buffer containing 20 mM Tris-HCl (pH 7.5), 20 mM KCl, 1 mM MgCl₂, and 1.5 mM dithiothreitol supplemented with 10 mM Na-phosphate (neutralized). In either case, aliquots were withdrawn at the appropriate times, quenched in 3 volumes of a buffer containing 90% deionized formamide, denatured by boiling, and separated on an 8% sequencing gel. Data were quantified by measuring the recoveries of substrate and partially digested (stalled) intermediates by using a phosphorimager.

RESULTS

RNA substrates of defined structure.

Figure 1 illustrates the structure of the RNA substrates used in this investigation. Three different classes of 3′ extensions were created. The first, denoted SLx (where x denotes the number of base pairs in the stem-loop structure), contains a six-residue 3′ extension, GGGAUC (Fig. 1a, top). The second, SLx-A, contains a 41-residue 3′ extension, GGGAUCC[A]₃₀UCUAG (Fig. 1a, middle), which mimics a typical bacterial poly(A) tail. The third class of substrate, SLx-R, contains a 36-residue extension corresponding to the polylinker of the vector 3′ to the BamHI site (5′-GGG AUCCUCUAGAGUCGACCUGCAGGCAUGCAAGCU). This extension is essentially random and duplicates the 3′ extension of a model substrate, t40B, used previously (6). Four different stem-loop structures, shown in Fig. 1b, were combined with each class of extension. Each stem-loop contains a four-residue loop (a GNRA tetraloop in one case) and a stem of 5, 7, 9, or 11 G-C base pairs (Fig. 1b).

The SLx and SLxA RNAs were predicted to fold as shown in Fig. 1 by using the program RNAdraw (23; http://rnadraw.base8.se/). RNAs SL5 and SL5A could also form weak alternative structures (not shown). RNAs in the SLxR class were also predicted to fold as designed but could form an additional imperfect stem-loop 3′ to the designed stem-loop (shown by the arrows in Fig. 1a). Some other structures could also be formed involving base pairing between the 5′ and 3′ extensions of the SLxR RNAs (not shown). Accordingly, structure mapping of SL11R RNA was employed to assess the extent to which alternative structures could form under conditions for exonuclease digestion. All C and G residues in the 3′ extension were accessible to RNase CL3 or T1, respectively, suggesting that the major fraction of SL11R exists in the form shown in Fig. 1a (data not shown).

Exonuclease digestion of SLx and SLxA RNAs.

RNAs in the SLx class mimic the products obtained after RNase E digestion of a typical mRNA substrate inasmuch as this endonuclease cleaves in single-stranded regions, often between stem-loop structures, thus leaving a short 3′ extension beyond the base of the stem-loop (9, 14, 20). Each of the RNA substrates (the multiple bands are due to 3′-terminal heterogeneity) was digested with RNase II (Fig. 2a) or PNPase (Fig. 2b), and the fraction of full-length substrate or intermediate products remaining was measured (Table 1). The three most stable substrates, SL11, SL9, and SL7, are almost completely resistant to either RNase II or PNPase although the same dilution of either enzyme was fully active against other substrates (see below). In contrast, SL5 RNA does undergo limited digestion by PNPase (12% of the substrate disappears in 10 min) but not by RNase II (Fig. 2 and Table 1). The relative resistance of SL5 RNA could be overcome by increasing the amount of either exonuclease in the assay by 10-fold, with the result that 61 and 72% of the SL5 RNA was digested by RNase II and PNPase, respectively, after 10 min of digestion (data not shown). We presume that this reflects the ability of the added enzyme to “capture” RNAs which have spontaneously melted (see the Discussion).

FIG. 2.

Digestion of SLx RNAs with 3′ exonucleases. Digestions were performed as described in Materials and Methods. Aliquots were removed at 0, 2.5, 5, and 10 min of digestion and denatured. RNAs were separated by electrophoresis under denaturing conditions, and the products were visualized by autoradiography or by phosphorimaging (see Materials and Methods). Shown are digestions with RNase II (a) and PNPase (b). Each panel is a separate gel. The schematic in the central margin illustrates the structure of the substrates.

TABLE 1.

Relative susceptibility of substrates to exoribonucleases^a

Substrate	% of substrate digested by enzyme
	RNase II	PNPase
SL5	1	12
SL5A	45	30
SL5R	35	44
SL7	0	0
SL7A	4	15
SL7R	17	22
SL9	0	4
SL9A	9	16
SL9R	16	15
SL11	2	5
SL11A	4	3
SL11R	6	0

^aLabeled RNA substrates (see the text and Fig. 1) were prepared and assayed for susceptibility to RNase II or PNPase as described in Materials and Methods. The data are the percentages of original substrate or stalled intermediate(s) completely converted to mononucleotides or limit oligonucleotides after 10 min of digestion. Each value is the average of three to six determinations. 

RNAs of the SLxA class containing a poly(A) tail were significantly more susceptible to exonuclease digestion than those of the SLx class (Fig. 3 and Table 1). These substrates were degraded in a two-step process: an initial rapid shortening from the 3′ end to yield a set of stalled intermediates was followed by a varied rate of disappearance of these intermediates. Each of the RNA substrates was shortened by either exonuclease, and approximately the same fraction of the full-length starting material, 50 to 70%, was attacked during the time of measurement in each case. RNAs SL11A, SL9A, and SL7A were converted quickly into relatively stable shorter intermediates denoted by the brackets in the center margin in Fig. 3. The intermediates from RNAs SL11A, SL9A, and SL7A accumulated almost quantitatively during the assay and displayed limited further susceptibility to either RNase II (Fig. 3a) or PNPase (Fig. 3b). Intermediates from SL7A RNA were largely resistant to RNase II but fourfold more susceptible to PNPase (Table 1). SL5A was susceptible to both enzymes, as from 30 to 45% of the starting material was converted to mononucleotides and limit oligonucleotides by RNase II or by PNPase, respectively, with the accumulation of only a faint ladder of intermediates (Fig. 3a, lanes 16 to 20).

FIG. 3.

Digestion of SLxA RNAs with 3′ exonucleases. Digestions were performed as described in Materials and Methods. Aliquots were removed after 0, 2.5, 5, 7.5 (a only), and 10 min of digestion, denatured, and resolved as described in the legend to Fig. 2. Shown are digestions with RNase II (a) and PNPase (b). Brackets in the margins point to intermediates which accumulate during digestion (see the text). Each panel is a different gel; therefore, the relative mobilities of products cannot be compared directly between panels. Lanes with sequence ladders are not shown. The schematic diagrams in the margin denote the structures of the substrate and intermediates. Lane M, undigested SL11A RNA.

The sizes of the various intermediates obtained from digestion of SLxA RNAs were determined by comparison to a ladder generated by partial alkaline hydrolysis or partial (random) T₁ RNase digestion of a 5′-end-labeled substrate. In general, each enzyme produced a distribution of three to five favored end points 3′ to a stem-loop (Table 2). The most prominent intermediate produced by RNase II from SL11A maps eight residues 3′ to the base of the stem-loop; others map between six and nine residues from its base (Fig. 3a, lanes 1 to 5; Table 2). A faster-moving doublet at the bottom of lanes 4 and 5 and a similarly sized band in lanes 17 to 20 of Fig. 3a map to the residue(s) 3′ to the base of the stem-loop. These bands could not be recovered reproducibly, however. The most prominent intermediate from PNPase digestion of SL11A also maps eight residues to the 3′ side of the stem-loop, with others at seven or nine residues (Fig. 3b, lanes 1 to 5; Table 2). A similar distribution of stalled intermediates was measured with the other SLxA substrates (Table 2), although faint shorter and longer intermediates were also observed (i.e., 5 or 10 residues 3′ to the base of the stem-loop).

TABLE 2.

Stalling points of exoribonucleases 3′ to stable stem-loops^a

Substrate	Stalling pointb
	RNase II	PNPase
SL5A	ND	ND
SL5R	ND	ND
SL7A	+6–+10	+7–+9
SL7R	+9–+11	ND
SL9A	+6–+9	+7–+9
SL9R	+6–+11	+7–+9
SL11A	+6–+9	+7–+9
SL11R	+8–+11	+8–+11

^aA stalled intermediate(s) arising from digestion with either RNase II or PNPase was separated on a sequencing gel and compared in size to ladders from partial alkaline or partial T₁ digests of the 5′-end-labeled substrates. 

^bResidue +1 is the first unpaired G residue 3′ to the stem-loop (see Fig. 1a). 

^cND, not determined (insufficient yields). 

Exonuclease digestion of the SLxR series of substrates.

Although we recognized that members of the SLxR series of substrates, designed as a control for the SLxA series, might prove problematic in view of their potential to engage in intra- and intermolecular base pairing through 5′ extensions, we examined their susceptibility to digestion under conditions identical to those used above. SL5R RNA was relatively susceptible to both RNase II and PNPase. At least 35% of the substrate disappeared totally within 10 min, and most of the starting material was converted to shorter intermediates (see Fig. 4a, lanes 16 to 20, and Table 1). In contrast, although full-length SL11R RNA was digestible by either exonuclease, as a significant fraction of the initial substrate disappeared during the assay, it was converted to relatively stable shorter intermediate products, much like SL11A. A minor RNase II product from digestion of SL11R RNA (arrowhead in Fig. 4a) represents removal of about three or four residues from the extreme 3′ end. This may represent stalling of RNase II 3′ to an imperfect 7-bp stem-loop formed by pairing between the sequence 5′-CCUCUAG and 5′-CUGCAGG (G-U and C-C mismatches are underlined; see also Fig. 1b). A similar minor product was also observed with SL9R in other experiments (not visible in Fig. 4a, lanes 6 to 10). A further set of major products shown by brackets in Fig. 4a, lanes 1 to 5, represents stalling 3′ to the 11-bp stem. In the assays of digestion by PNPase, small amounts of an intermediate shortened by just three or four residues, similar to that obtained with RNase II, appeared transiently (Fig. 4b, lanes 1 to 5) but did not accumulate. Rather, over 80% of the initial substrate was converted further to stable intermediates corresponding to stalling 8 to 11 residues 3′ to the 11-bp stem-loop. Increasing the concentration of RNase II by up to 10-fold or that of PNPase by up to 5-fold enhanced the initial rate of shortening of SL11R RNA, as would be expected. Neither increase altered the extent of conversion of SL11R RNA to stalled intermediates or the size of the latter (data not shown). This differs from the response of SL5 RNA and presumably reflects the high stability of the stem-loop in SL11R. SL7R and SL9R RNAs exhibited behavior between that of SL5R and SL11R.

FIG. 4.

Digestion of SLxR RNAs with 3′ exonucleases. Digestions were performed as described in Materials and Methods. Aliquots were removed at 0, 2.5, 5, 7.5, and 10 min of digestion, denatured, and resolved as described in the legend to Fig. 2. Shown are digestions with RNase II (a) and PNPase (b). Each panel is a separate gel. Lanes with sequence ladders are not shown. Brackets in the margins mark the positions of intermediates presumed to be the result of stalling of the enzyme. The arrowheads show the position of a transient intermediate (see the text). The schematic diagrams in the central margin denote the structures of the substrate and intermediates (see the text).

DISCUSSION

Functional similarities between RNase II and PNPase.

To our knowledge, this is the first systematic comparison of the properties of RNase II and PNPase in vitro, although we and other investigators have examined these enzymes' activities against a single or limited number of substrates (6–8, 17, 18, 24). We chose conditions in which each enzyme exhibited comparable exoribonuclease activity and in which the activity was limiting. To a large extent, the determinants of these enzymes' abilities to attack different substrates are quite similar. Neither enzyme is able to attack RNA stem-loops containing a 3′-terminal extension of six residues, in agreement with previous data in vitro (6) and in vivo (3, 28). However, comparably stable RNAs in the SLxA or SLxR class are readily shortened by both enzymes. The partial removal of the 3′ extension on these substrates by both enzymes shows that there is no initial barrier to initiation of exonucleolytic attack; rather, the barrier to continued digestion is the internal stem-loop itself. These observations are best explained by a model in which either exonuclease requires a single-stranded RNA target larger than six residues for initial binding. PNPase is very inefficient against short oligomers (K_m values from 50 to 250 mM), but becomes highly efficient (K_m of 10 nM) for oligomers greater than 10 residues (16). Likewise, RNase II loses processivity on poly(A) containing fewer than 10 to 15 residues (4). Whether stalling (and subsequent dissociation) or complete digestion occurs will depend on the balance between the rate of enzyme dissociation and the rate of transient melting of weaker stem-loop barriers, such as in SL5A and SL7A. The ability of increased concentrations of RNase II or PNPase to overcome the resistance of SL5 to decay may simply reflect the greater likelihood of either enzyme binding to a partially or fully melted substrate molecule before the latter refolds (“substrate capture”).

The ability of either exonuclease, but especially of RNase II, the major 3′ exonuclease in crude extracts (12), to remove poly(A) extensions would indirectly impede exonuclease action against structured RNAs (3, 6, 28). This property is likely the basis of the observation that RNase II can stabilize RNA-OUT and fragments of the rpsO mRNA (3, 28). At first glance, such behavior would seem to promote futile cycles of polyadenylation and deadenylation. We believe that it serves to drive mRNA decay into an RNase E-dependent mode, promoting 5′→3′ decay (9).

Previous investigations have reported that PNPase will stall six residues 3′ to a stable RNA stem-loop (24) whereas RNase II stalls only four residues away (17, 24). In contrast, we observed only modest differences between situations in which the two exoribonucleases cease digestion before encountering a stable secondary structure. Our data for SL11A RNA show that there is a distribution of stalling points with the most frequent of these occurring somewhat more 3′ than reported by others (eight residues from the base of the stem for both RNase II and PNPase). We believe that the differences among published reports must reflect variations in substrate sequence and composition. In this regard, the SL RNA substrates contain stem-loops which are composed of G-C base pairs exclusively; in addition, the three first residues 3′ to the base of the stem-loop are Gs.

Is the behavior of the exoribonucleases in vitro consistent with mRNA decay in vivo?

Apart from repetitive extragenic palindrome (REP) sequences and rho-independent terminators, perfectly matched stem-loops containing more than six contiguous base pairs ought to occur relatively infrequently in most mRNAs (frequency, <1 in 4,056 residues). Imperfect stem-loops would, of course, occur much more often. Moreover, all such structures would contain roughly equal frequencies of A-U and G-C base pairs, unlike the more stable model RNAs used here. Thus, it is unlikely that either exoribonuclease would encounter many highly stable, internal barriers in most mRNA fragments created by RNase E cleavage. Rather, new 3′ ends will most often be generated in regions which are unstructured or contain weak stem-loops which are in rapid equilibrium with their alternatively folded or unfolded counterparts. The structure of such cleavage products would permit access by either exoribonuclease as long as more than 10 to 12 residues at their extremities were or could become single-stranded. An example of this situation occurs in the rpsT mRNA. A major RNase E cleavage site occurs five residues 3′ to a moderately stable (ΔG = −2.7 kCal/mol) stem-loop (stem-loop IV; see reference 20). This structure is not a barrier to RNase II or to PNPase in vitro or apparently in vivo (7, 19). To this extent, these two enzymes are functionally redundant. In those cases where a stable stem-loop does occur less than 10 to 12 residues 5′ to an RNase E cleavage site, our data show that oligoadenylation can restore the accessibility of the cleavage product to both exoribonucleases (7, 8).

The two 3′ exoribonucleases do, however, differ significantly in one regard. PNPase spontaneously forms a complex with RNase E (5, 10, 27, 29, 32), in which form its activity can be stimulated by RhlB, a DEAD-box RNA helicase. This relatively recently appreciated property of PNPase allows it to catalyze the degradation of highly folded RNAs, including REP sequences or rho-independent terminators (10, 29). This could explain the reports of differential degradation of structured RNAs dependent on PNPase (3, 7, 15, 19). In contrast, RNase II does not bind to components of the degradosome and cannot be activated by RhlB, thereby limiting its ability to degrade RNAs containing stable stem-loops (10).